Files
yolov26_3d/ultralytics/nn/modules/head.py
2026-06-24 09:35:46 +08:00

2104 lines
92 KiB
Python
Executable File
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
"""Model head modules."""
from __future__ import annotations
import copy
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.init import constant_, xavier_uniform_
from ultralytics.utils import NOT_MACOS14
from ultralytics.utils.tal import dist2bbox, dist2rbox, make_anchors
from ultralytics.utils.torch_utils import TORCH_1_11, fuse_conv_and_bn, smart_inference_mode
from .block import DFL, SAVPE, BNContrastiveHead, ContrastiveHead, Proto, Proto26, RealNVP, Residual, SwiGLUFFN
from .conv import Conv, DWConv
from .transformer import MLP, DeformableTransformerDecoder, DeformableTransformerDecoderLayer
from .utils import bias_init_with_prob, linear_init
__all__ = "OBB", "Classify", "Detect", "Detect3D", "Pose", "RTDETRDecoder", "Segment", "YOLOEDetect", "YOLOESegment", "v10Detect"
class Detect(nn.Module):
"""YOLO Detect head for object detection models.
This class implements the detection head used in YOLO models for predicting bounding boxes and class probabilities.
It supports both training and inference modes, with optional end-to-end detection capabilities.
Attributes:
dynamic (bool): Force grid reconstruction.
export (bool): Export mode flag.
format (str): Export format.
end2end (bool): End-to-end detection mode.
max_det (int): Maximum detections per image.
shape (tuple): Input shape.
anchors (torch.Tensor): Anchor points.
strides (torch.Tensor): Feature map strides.
legacy (bool): Backward compatibility for v3/v5/v8/v9/v11 models.
xyxy (bool): Output format, xyxy or xywh.
nc (int): Number of classes.
nl (int): Number of detection layers.
reg_max (int): DFL channels.
no (int): Number of outputs per anchor.
stride (torch.Tensor): Strides computed during build.
cv2 (nn.ModuleList): Convolution layers for box regression.
cv3 (nn.ModuleList): Convolution layers for classification.
dfl (nn.Module): Distribution Focal Loss layer.
one2one_cv2 (nn.ModuleList): One-to-one convolution layers for box regression.
one2one_cv3 (nn.ModuleList): One-to-one convolution layers for classification.
Methods:
forward: Perform forward pass and return predictions.
bias_init: Initialize detection head biases.
decode_bboxes: Decode bounding boxes from predictions.
postprocess: Post-process model predictions.
Examples:
Create a detection head for 80 classes
>>> detect = Detect(nc=80, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = detect(x)
"""
dynamic = False # force grid reconstruction
export = False # export mode
format = None # export format
max_det = 300 # max_det
agnostic_nms = False
shape = None
anchors = torch.empty(0) # init
strides = torch.empty(0) # init
legacy = False # backward compatibility for v3/v5/v8/v9 models
xyxy = False # xyxy or xywh output
def __init__(self, nc: int = 80, reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize the YOLO detection layer with specified number of classes and channels.
Args:
nc (int): Number of classes.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__()
self.nc = nc # number of classes
self.nl = len(ch) # number of detection layers
self.reg_max = reg_max # DFL channels (ch[0] // 16 to scale 4/8/12/16/20 for n/s/m/l/x)
self.no = nc + self.reg_max * 4 # number of outputs per anchor
self.stride = torch.zeros(self.nl) # strides computed during build
c2, c3 = max((16, ch[0] // 4, self.reg_max * 4)), max(ch[0], min(self.nc, 100)) # channels
self.cv2 = nn.ModuleList(
nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch
)
self.cv3 = (
nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch)
if self.legacy
else nn.ModuleList(
nn.Sequential(
nn.Sequential(DWConv(x, x, 3), Conv(x, c3, 1)),
nn.Sequential(DWConv(c3, c3, 3), Conv(c3, c3, 1)),
nn.Conv2d(c3, self.nc, 1),
)
for x in ch
)
)
self.cv_diff = nn.ModuleList(
nn.Sequential(Conv(x, x, 1), Conv(x, x, 1), nn.Conv2d(x, 1, 1)) for x in ch
)
self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity()
if end2end:
self.one2one_cv2 = copy.deepcopy(self.cv2)
self.one2one_cv3 = copy.deepcopy(self.cv3)
self.one2one_cv_diff = copy.deepcopy(self.cv_diff)
@property
def one2many(self):
"""Returns the one-to-many head components, here for v3/v5/v8/v9/v11 backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, diff_head=self.cv_diff)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(box_head=self.one2one_cv2, cls_head=self.one2one_cv3, diff_head=self.one2one_cv_diff)
@property
def end2end(self):
"""Checks if the model has one2one for v3/v5/v8/v9/v11 backward compatibility."""
return getattr(self, "_end2end", True) and hasattr(self, "one2one")
@end2end.setter
def end2end(self, value):
"""Override the end-to-end detection mode."""
self._end2end = value
def forward_head(
self,
x: list[torch.Tensor],
box_head: torch.nn.Module = None,
cls_head: torch.nn.Module = None,
diff_head: torch.nn.Module = None,
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes and class probabilities."""
if box_head is None or cls_head is None: # for fused inference
return dict()
bs = x[0].shape[0] # batch size
boxes = torch.cat([box_head[i](x[i]).view(bs, 4 * self.reg_max, -1) for i in range(self.nl)], dim=-1)
scores = torch.cat([cls_head[i](x[i]).view(bs, self.nc, -1) for i in range(self.nl)], dim=-1)
preds = dict(boxes=boxes, scores=scores, feats=x)
if diff_head is not None:
preds["preds_diff"] = torch.cat([diff_head[i](x[i]).view(bs, 1, -1) for i in range(self.nl)], dim=-1)
return preds
def forward(
self, x: list[torch.Tensor]
) -> dict[str, torch.Tensor] | torch.Tensor | tuple[torch.Tensor, dict[str, torch.Tensor]]:
"""Concatenates and returns predicted bounding boxes and class probabilities."""
preds = self.forward_head(x, **self.one2many)
if self.end2end:
x_detach = [xi.detach() for xi in x]
one2one = self.forward_head(x_detach, **self.one2one)
preds = {"one2many": preds, "one2one": one2one}
if self.training:
return preds
y = self._inference(preds["one2one"] if self.end2end else preds)
if self.end2end:
y = self.postprocess(y.permute(0, 2, 1))
preds_diff = preds["one2one"].get("preds_diff")
if preds_diff is not None and hasattr(self, "_last_topk_idx"):
preds["one2one"]["preds_diff_selected"] = self._select_topk_branch(preds_diff, self._last_topk_idx)
return y if self.export else (y, preds)
def _inference(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Decode predicted bounding boxes and class probabilities based on multiple-level feature maps.
Args:
x (dict[str, torch.Tensor]): Dictionary of predictions from detection layers.
Returns:
(torch.Tensor): Concatenated tensor of decoded bounding boxes and class probabilities.
"""
# Inference path
dbox = self._get_decode_boxes(x)
return torch.cat((dbox, x["scores"].sigmoid()), 1)
def _get_decode_boxes(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Get decoded boxes based on anchors and strides."""
shape = x["feats"][0].shape # BCHW
if self.dynamic or self.shape != shape:
self.anchors, self.strides = (a.transpose(0, 1) for a in make_anchors(x["feats"], self.stride, 0.5))
self.shape = shape
dbox = self.decode_bboxes(self.dfl(x["boxes"]), self.anchors.unsqueeze(0)) * self.strides
return dbox
def bias_init(self):
"""Initialize Detect() biases, WARNING: requires stride availability."""
for i, (a, b) in enumerate(zip(self.one2many["box_head"], self.one2many["cls_head"])): # from
a[-1].bias.data[:] = 2.0 # box
b[-1].bias.data[: self.nc] = math.log(
5 / self.nc / (640 / self.stride[i]) ** 2
) # cls (.01 objects, 80 classes, 640 img)
if self.end2end:
for i, (a, b) in enumerate(zip(self.one2one["box_head"], self.one2one["cls_head"])): # from
a[-1].bias.data[:] = 2.0 # box
b[-1].bias.data[: self.nc] = math.log(
5 / self.nc / (640 / self.stride[i]) ** 2
) # cls (.01 objects, 80 classes, 640 img)
def decode_bboxes(self, bboxes: torch.Tensor, anchors: torch.Tensor, xywh: bool = True) -> torch.Tensor:
"""Decode bounding boxes from predictions."""
return dist2bbox(
bboxes,
anchors,
xywh=xywh and not self.end2end and not self.xyxy,
dim=1,
)
def postprocess(self, preds: torch.Tensor) -> torch.Tensor:
"""Post-processes YOLO model predictions.
Args:
preds (torch.Tensor): Raw predictions with shape (batch_size, num_anchors, 4 + nc) with last dimension
format [x1, y1, x2, y2, class_probs].
Returns:
(torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 6) and last
dimension format [x1, y1, x2, y2, max_class_prob, class_index].
"""
boxes, scores = preds.split([4, self.nc], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
self._last_topk_idx = idx
return torch.cat([boxes, scores, conf], dim=-1)
@staticmethod
def _select_topk_branch(preds_branch, idx):
"""Select per-anchor branch predictions aligned with per-sample top-k detections."""
idx_branch = idx.squeeze(-1).unsqueeze(1).expand(-1, preds_branch.shape[1], -1)
return preds_branch.gather(2, idx_branch).permute(0, 2, 1)
def get_topk_index(self, scores: torch.Tensor, max_det: int) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
"""Get top-k indices from scores.
Args:
scores (torch.Tensor): Scores tensor with shape (batch_size, num_anchors, num_classes).
max_det (int): Maximum detections per image.
Returns:
(torch.Tensor, torch.Tensor, torch.Tensor): Top scores, class indices, and filtered indices.
"""
batch_size, anchors, nc = scores.shape # i.e. shape(16,8400,84)
# Use max_det directly during export for TensorRT compatibility (requires k to be constant),
# otherwise use min(max_det, anchors) for safety with small inputs during Python inference
k = max_det if self.export else min(max_det, anchors)
if self.agnostic_nms:
scores, labels = scores.max(dim=-1, keepdim=True)
scores, indices = scores.topk(k, dim=1)
labels = labels.gather(1, indices)
return scores, labels, indices
ori_index = scores.max(dim=-1)[0].topk(k)[1].unsqueeze(-1)
scores = scores.gather(dim=1, index=ori_index.repeat(1, 1, nc))
scores, index = scores.flatten(1).topk(k)
# Original implementation kept for reference. It exports `aten::index` and `%`,
# which may become ONNX `Gather` chains and `Mod`, both unfriendly to some chip toolchains.
idx = ori_index[torch.arange(batch_size)[..., None], index // nc] # original index
return scores[..., None], (index % nc)[..., None].float(), idx
# flat_anchor_index = ori_index.expand(-1, -1, nc).reshape(batch_size, -1, 1)
# idx = flat_anchor_index.gather(1, index.unsqueeze(-1)) # original anchor index
# class_lookup = torch.arange(nc, device=index.device, dtype=index.dtype).view(1, 1, nc)
# class_lookup = class_lookup.expand(batch_size, k, nc).reshape(batch_size, -1)
# class_index = class_lookup.gather(1, index)
# return scores[..., None], class_index[..., None].float(), idx
def fuse(self) -> None:
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = self.cv_diff = None
class Detect3D(Detect):
"""YOLO 3D detection head extending Detect with 3D prediction branches.
Adds cv4 branches for 3D predictions (depth, UV offsets, dimensions, yaw, face visibility)
alongside existing cv2 (box) and cv3 (cls) branches. Follows the Pose head pattern.
3D output format (41 channels per anchor):
- Channels 0-5: Front face (z3d, u_offset, v_offset, h, w, visible_score)
- Channels 6-11: Rear face (z3d, u_offset, v_offset, h, w, visible_score)
- Channels 12-17: Left face (z3d, u_offset, v_offset, l, h, visible_score)
- Channels 18-23: Right face (z3d, u_offset, v_offset, l, h, visible_score)
- Channels 24-40: Whole 3D box
- 24: z3d (depth in meters)
- 25-26: u_offset, v_offset (bounded grid offsets)
- 27-29: l, h, w (dimensions in meters)
- 30-33: yaw class logits (4 orientation bins)
- 34-37: yaw residual sine values for the 4 orientation bins
- 38-40: cut class logits (3 classes: normal/cut_in/cut_out)
After denormalization in forward pass, outputs are in physical units:
- z3d: meters, UV: grid cells [-3.5, 4.5], size: meters, yaw_reg: sin(delta) in [-1, 1].
Attributes:
no_3d (int): Number of 3D output channels per anchor (41).
cv4 (nn.ModuleList): 3D prediction conv branches.
cv6 (nn.ModuleList): Additional 3D prediction conv branches for fake classes.
norm_scales_3d (dict): Normalization scales set by trainer after model creation.
"""
no_3d = 41 # 4 faces × 6 + 17 whole-box
edge_point_count = 5
edge_point_dims = 3 # du, dv, z per sampled point
edge_face_dims = edge_point_count * edge_point_dims
no_edge = 4 * edge_face_dims # 4 faces × 5 sampled points × (du, dv, z)
uv_range = 16.0 # decoded UV offset range in grid cells
uv_shift = 8.0 # centered so raw=0 -> decoded offset 0
def __init__(self, nc=80, reg_max=16, end2end=False, ch=()):
"""Initialize Detect3D head with 3D prediction branches."""
super().__init__(nc, reg_max, end2end, ch)
self.cv4 = nn.ModuleList(
nn.Sequential(Conv(x, x, 1), Conv(x, x, 1), nn.Conv2d(x, self.no_3d, 1)) for x in ch
)
self.cv5 = nn.ModuleList(
nn.Sequential(Conv(x, x, 1), Conv(x, x, 1), nn.Conv2d(x, self.no_edge, 1)) for x in ch
)
self.cv6 = nn.ModuleList(
nn.Sequential(Conv(x, x, 1), Conv(x, x, 1), nn.Conv2d(x, self.no_3d, 1)) for x in ch
)
if end2end:
self.one2one_cv4 = copy.deepcopy(self.cv4)
self.one2one_cv5 = copy.deepcopy(self.cv5)
self.one2one_cv6 = copy.deepcopy(self.cv6)
# Set by trainer after model creation via model.norm_scales_3d
self.norm_scales_3d = {}
def bias_init(self):
"""Initialize 2D detect biases plus stable priors for the 3D regression heads."""
super().bias_init()
def _init_3d_branch(branches):
for branch in branches:
head = branch[-1]
if not isinstance(head, nn.Conv2d):
continue
# Start from explicit priors instead of random Conv2d init.
head.weight.data.zero_()
head.bias.data.zero_()
# z raw bias = 0 -> denorm to z_offset (dataset mean depth prior).
for ch in (0, 6, 12, 18, 24):
head.bias.data[ch] = 0.0
# With the symmetric UV decoder, raw 0 already maps to 0 grid-cell offset.
for ch_start in (1, 7, 13, 19, 25):
head.bias.data[ch_start : ch_start + 2] = 0.0
# size raw bias = 0 -> denorm to size_offset prior.
for ch_start in (3, 9, 15, 21):
head.bias.data[ch_start : ch_start + 2] = 0.0
head.bias.data[27:30] = 0.0
# Yaw, cut logits, and face visibility start neutral.
head.bias.data[30:41] = 0.0
def _init_edge_branch(branches):
for branch in branches:
head = branch[-1]
if not isinstance(head, nn.Conv2d):
continue
head.weight.data.zero_()
head.bias.data.zero_()
for face_start in range(0, self.no_edge, self.edge_face_dims):
for point_start in range(face_start, face_start + self.edge_face_dims, self.edge_point_dims):
head.bias.data[point_start : point_start + 2] = 0.0
head.bias.data[point_start + 2] = 0.0
_init_3d_branch(self.cv4)
_init_edge_branch(self.cv5)
_init_3d_branch(self.cv6)
if self.end2end:
_init_3d_branch(self.one2one_cv4)
_init_edge_branch(self.one2one_cv5)
_init_3d_branch(self.one2one_cv6)
@property
def one2many(self):
"""Returns the one-to-many head components including 3D branch."""
return dict(
box_head=self.cv2,
cls_head=self.cv3,
head_3d=self.cv4,
edge_head=self.cv5,
fake_head_3d=self.cv6,
diff_head=self.cv_diff,
)
@property
def one2one(self):
"""Returns the one-to-one head components including 3D branch."""
return dict(
box_head=self.one2one_cv2,
cls_head=self.one2one_cv3,
head_3d=self.one2one_cv4,
edge_head=self.one2one_cv5,
fake_head_3d=self.one2one_cv6,
diff_head=self.one2one_cv_diff,
)
def forward_head(
self, x, box_head=None, cls_head=None, head_3d=None, edge_head=None, fake_head_3d=None, diff_head=None
):
"""Forward pass including 3D predictions with inverse label normalization."""
preds = super().forward_head(x, box_head, cls_head, diff_head=diff_head)
bs = x[0].shape[0]
if head_3d is not None:
raw_3d = torch.cat([head_3d[i](x[i]).view(bs, self.no_3d, -1) for i in range(self.nl)], dim=-1)
preds["preds_3d"] = self._denorm_3d(raw_3d)
if fake_head_3d is not None:
raw_fake_3d = torch.cat([fake_head_3d[i](x[i]).view(bs, self.no_3d, -1) for i in range(self.nl)], dim=-1)
preds["preds_3d_fake"] = self._denorm_3d(raw_fake_3d)
if edge_head is not None:
raw_edge = torch.cat([edge_head[i](x[i]).view(bs, self.no_edge, -1) for i in range(self.nl)], dim=-1)
preds["preds_edge"] = self._denorm_edge(raw_edge)
return preds
def _denorm_3d(self, p3d):
"""Denormalize raw 3D predictions to physical units.
Matches yolov5-3d Detect2D3D.denormalize_3d_predictions():
- z3d: raw * z3d_scale + z3d_offset → meters
- UV offsets: sigmoid(raw) * uv_range - uv_shift → bounded grid offset [-8, 8]
- size: raw * size_scale + size_offset → meters
- yaw residuals: tanh(raw) → sin(delta) in [-1, 1]
- yaw_cls, cut_cls: kept as raw logits
- face_vis: kept as raw score
Args:
p3d (Tensor): Raw predictions [B, 41, A] (batch, channels, anchors).
Returns:
Tensor: Denormalized predictions [B, 41, A].
"""
ns = self.norm_scales_3d
z_scale = ns.get("z3d_scale", 1.0)
z_offset = ns.get("z3d_offset", 0.0)
s_scale = ns.get("size_scale", 1.0)
s_offset = ns.get("size_offset", 0.0)
p = p3d.clone()
# z3d → meters: face channels 0, 6, 12, 18 and whole channel 24
for ch in (0, 6, 12, 18, 24):
p[:, ch] = p[:, ch] * z_scale + z_offset
# UV offsets → bounded grid offset: face 1-2, 7-8, 13-14, 19-20 and whole 25-26
for ch_start in (1, 7, 13, 19, 25):
p[:, ch_start : ch_start + 2] = p[:, ch_start : ch_start + 2].sigmoid() * self.uv_range - self.uv_shift
# size → meters: face 3-4, 9-10, 15-16, 21-22 and whole 27-29
for ch_start in (3, 9, 15, 21):
p[:, ch_start : ch_start + 2] = p[:, ch_start : ch_start + 2] * s_scale + s_offset
p[:, 27:30] = p[:, 27:30] * s_scale + s_offset
# yaw residual sine → [-1, 1]: channels 34-37
p[:, 34:38] = p[:, 34:38].tanh()
return p
def _denorm_edge(self, p_edge):
"""Denormalize raw visible-face edge predictions to grid-space UV offsets and metric depths."""
ns = self.norm_scales_3d
z_scale = ns.get("z3d_scale", 1.0)
z_offset = ns.get("z3d_offset", 0.0)
p = p_edge.clone()
for face_start in range(0, self.no_edge, self.edge_face_dims):
for point_start in range(face_start, face_start + self.edge_face_dims, self.edge_point_dims):
p[:, point_start : point_start + 2] = (
p[:, point_start : point_start + 2].sigmoid() * self.uv_range - self.uv_shift
)
p[:, point_start + 2] = p[:, point_start + 2] * z_scale + z_offset
return p
def postprocess(self, preds):
"""Post-process predictions, storing top-k indices for 3D selection."""
boxes, scores = preds.split([4, self.nc], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
self._last_topk_idx = idx # (B, k, 1) — anchor indices for 3D selection
return torch.cat([boxes, scores, conf], dim=-1)
def _select_topk_3d_metadata(self, preds_3d, idx):
"""Select 3D predictions and anchor metadata aligned with per-sample top-k detections."""
# Original axis=2 implementation kept for reference.
idx_3d = idx.squeeze(-1).unsqueeze(1).expand(-1, self.no_3d, -1)
preds_3d_selected = preds_3d.gather(2, idx_3d).permute(0, 2, 1)
idx_flat = idx.squeeze(-1) # (B, k)
# preds_3d_bac = preds_3d.permute(0, 2, 1)
# idx_3d = idx_flat.unsqueeze(-1).expand(-1, -1, self.no_3d)
# preds_3d_selected = preds_3d_bac.gather(1, idx_3d)
# Original axis=2 implementation kept for reference.
anchor_idx = idx_flat.unsqueeze(1).expand(-1, self.anchors.shape[0], -1)
anchors_selected = self.anchors.unsqueeze(0).expand(idx_flat.shape[0], -1, -1).gather(2, anchor_idx)
# anchors_ba2 = self.anchors.transpose(0, 1).unsqueeze(0).expand(idx_flat.shape[0], -1, -1)
# anchor_idx = idx_flat.unsqueeze(-1).expand(-1, -1, anchors_ba2.shape[-1])
# anchors_selected = anchors_ba2.gather(1, anchor_idx).permute(0, 2, 1)
strides_selected = self.strides.expand(idx_flat.shape[0], -1).gather(1, idx_flat)
return preds_3d_selected, anchors_selected, strides_selected
@staticmethod
def _select_topk_branch(preds_branch, idx):
"""Select per-anchor branch predictions aligned with per-sample top-k detections."""
# Original axis=2 implementation kept for reference.
idx_branch = idx.squeeze(-1).unsqueeze(1).expand(-1, preds_branch.shape[1], -1)
return preds_branch.gather(2, idx_branch).permute(0, 2, 1)
# idx_flat = idx.squeeze(-1)
# preds_branch_bac = preds_branch.permute(0, 2, 1)
# idx_branch = idx_flat.unsqueeze(-1).expand(-1, -1, preds_branch.shape[1])
# return preds_branch_bac.gather(1, idx_branch)
def forward(self, x):
"""Forward pass with 3D prediction selection in eval mode."""
preds = self.forward_head(x, **self.one2many)
if self.end2end:
x_detach = [xi.detach() for xi in x]
one2one = self.forward_head(x_detach, **self.one2one)
preds = {"one2many": preds, "one2one": one2one}
if self.training:
return preds
y = self._inference(preds["one2one"] if self.end2end else preds)
if self.end2end:
y = self.postprocess(y.permute(0, 2, 1))
# Select corresponding 3D predictions using stored top-k indices
preds_3d = preds["one2one"].get("preds_3d")
if preds_3d is not None and hasattr(self, "_last_topk_idx"):
(
preds["one2one"]["preds_3d_selected"],
preds["one2one"]["anchors_selected"],
preds["one2one"]["strides_selected"],
) = self._select_topk_3d_metadata(preds_3d, self._last_topk_idx)
preds_edge = preds["one2one"].get("preds_edge")
if preds_edge is not None and hasattr(self, "_last_topk_idx"):
preds["one2one"]["preds_edge_selected"] = self._select_topk_branch(preds_edge, self._last_topk_idx)
preds_3d_fake = preds["one2one"].get("preds_3d_fake")
if preds_3d_fake is not None and hasattr(self, "_last_topk_idx"):
preds["one2one"]["preds_3d_fake_selected"] = self._select_topk_branch(
preds_3d_fake, self._last_topk_idx
)
preds_diff = preds["one2one"].get("preds_diff")
if preds_diff is not None and hasattr(self, "_last_topk_idx"):
preds["one2one"]["preds_diff_selected"] = self._select_topk_branch(preds_diff, self._last_topk_idx)
return y if self.export else (y, preds)
def _inference(self, x):
"""Decode predicted bounding boxes and class probabilities (3D not used at inference)."""
return super()._inference(x)
def fuse(self):
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = self.cv4 = self.cv5 = self.cv6 = self.cv_diff = None
class Segment(Detect):
"""YOLO Segment head for segmentation models.
This class extends the Detect head to include mask prediction capabilities for instance segmentation tasks.
Attributes:
nm (int): Number of masks.
npr (int): Number of protos.
proto (Proto): Prototype generation module.
cv4 (nn.ModuleList): Convolution layers for mask coefficients.
Methods:
forward: Return model outputs and mask coefficients.
Examples:
Create a segmentation head
>>> segment = Segment(nc=80, nm=32, npr=256, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = segment(x)
"""
def __init__(self, nc: int = 80, nm: int = 32, npr: int = 256, reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize the YOLO model attributes such as the number of masks, prototypes, and the convolution layers.
Args:
nc (int): Number of classes.
nm (int): Number of masks.
npr (int): Number of protos.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, reg_max, end2end, ch)
self.nm = nm # number of masks
self.npr = npr # number of protos
self.proto = Proto(ch[0], self.npr, self.nm) # protos
c4 = max(ch[0] // 4, self.nm)
self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, self.nm, 1)) for x in ch)
if end2end:
self.one2one_cv4 = copy.deepcopy(self.cv4)
@property
def one2many(self):
"""Returns the one-to-many head components, here for backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, mask_head=self.cv4)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(box_head=self.one2one_cv2, cls_head=self.one2one_cv3, mask_head=self.one2one_cv4)
def forward(self, x: list[torch.Tensor]) -> tuple | list[torch.Tensor] | dict[str, torch.Tensor]:
"""Return model outputs and mask coefficients if training, otherwise return outputs and mask coefficients."""
outputs = super().forward(x)
preds = outputs[1] if isinstance(outputs, tuple) else outputs
proto = self.proto(x[0]) # mask protos
if isinstance(preds, dict): # training and validating during training
if self.end2end:
preds["one2many"]["proto"] = proto
preds["one2one"]["proto"] = proto.detach()
else:
preds["proto"] = proto
if self.training:
return preds
return (outputs, proto) if self.export else ((outputs[0], proto), preds)
def _inference(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Decode predicted bounding boxes and class probabilities, concatenated with mask coefficients."""
preds = super()._inference(x)
return torch.cat([preds, x["mask_coefficient"]], dim=1)
def forward_head(
self, x: list[torch.Tensor], box_head: torch.nn.Module, cls_head: torch.nn.Module, mask_head: torch.nn.Module
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and mask coefficients."""
preds = super().forward_head(x, box_head, cls_head)
if mask_head is not None:
bs = x[0].shape[0] # batch size
preds["mask_coefficient"] = torch.cat([mask_head[i](x[i]).view(bs, self.nm, -1) for i in range(self.nl)], 2)
return preds
def postprocess(self, preds: torch.Tensor) -> torch.Tensor:
"""Post-process YOLO model predictions.
Args:
preds (torch.Tensor): Raw predictions with shape (batch_size, num_anchors, 4 + nc + nm) with last dimension
format [x1, y1, x2, y2, class_probs, mask_coefficient].
Returns:
(torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 6 + nm) and last
dimension format [x1, y1, x2, y2, max_class_prob, class_index, mask_coefficient].
"""
boxes, scores, mask_coefficient = preds.split([4, self.nc, self.nm], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
mask_coefficient = mask_coefficient.gather(dim=1, index=idx.repeat(1, 1, self.nm))
return torch.cat([boxes, scores, conf, mask_coefficient], dim=-1)
def fuse(self) -> None:
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = self.cv4 = None
class Segment26(Segment):
"""YOLO26 Segment head for segmentation models.
This class extends the Segment head with Proto26 for mask prediction in instance segmentation tasks.
Attributes:
nm (int): Number of masks.
npr (int): Number of protos.
proto (Proto26): Prototype generation module.
cv4 (nn.ModuleList): Convolution layers for mask coefficients.
Methods:
forward: Return model outputs and mask coefficients.
Examples:
Create a segmentation head
>>> segment = Segment26(nc=80, nm=32, npr=256, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = segment(x)
"""
def __init__(self, nc: int = 80, nm: int = 32, npr: int = 256, reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize the YOLO model attributes such as the number of masks, prototypes, and the convolution layers.
Args:
nc (int): Number of classes.
nm (int): Number of masks.
npr (int): Number of protos.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, nm, npr, reg_max, end2end, ch)
self.proto = Proto26(ch, self.npr, self.nm, nc) # protos
def forward(self, x: list[torch.Tensor]) -> tuple | list[torch.Tensor] | dict[str, torch.Tensor]:
"""Return model outputs and mask coefficients if training, otherwise return outputs and mask coefficients."""
outputs = Detect.forward(self, x)
preds = outputs[1] if isinstance(outputs, tuple) else outputs
proto = self.proto(x) # mask protos
if isinstance(preds, dict): # training and validating during training
if self.end2end:
preds["one2many"]["proto"] = proto
preds["one2one"]["proto"] = (
tuple(p.detach() for p in proto) if isinstance(proto, tuple) else proto.detach()
)
else:
preds["proto"] = proto
if self.training:
return preds
return (outputs, proto) if self.export else ((outputs[0], proto), preds)
def fuse(self) -> None:
"""Remove the one2many head and extra part of proto module for inference optimization."""
super().fuse()
if hasattr(self.proto, "fuse"):
self.proto.fuse()
class OBB(Detect):
"""YOLO OBB detection head for detection with rotation models.
This class extends the Detect head to include oriented bounding box prediction with rotation angles.
Attributes:
ne (int): Number of extra parameters.
cv4 (nn.ModuleList): Convolution layers for angle prediction.
angle (torch.Tensor): Predicted rotation angles.
Methods:
forward: Concatenate and return predicted bounding boxes and class probabilities.
decode_bboxes: Decode rotated bounding boxes.
Examples:
Create an OBB detection head
>>> obb = OBB(nc=80, ne=1, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = obb(x)
"""
def __init__(self, nc: int = 80, ne: int = 1, reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize OBB with number of classes `nc` and layer channels `ch`.
Args:
nc (int): Number of classes.
ne (int): Number of extra parameters.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, reg_max, end2end, ch)
self.ne = ne # number of extra parameters
c4 = max(ch[0] // 4, self.ne)
self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, self.ne, 1)) for x in ch)
if end2end:
self.one2one_cv4 = copy.deepcopy(self.cv4)
@property
def one2many(self):
"""Returns the one-to-many head components, here for backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, angle_head=self.cv4)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(box_head=self.one2one_cv2, cls_head=self.one2one_cv3, angle_head=self.one2one_cv4)
def _inference(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Decode predicted bounding boxes and class probabilities, concatenated with rotation angles."""
# For decode_bboxes convenience
self.angle = x["angle"] # TODO: need to test obb
preds = super()._inference(x)
return torch.cat([preds, x["angle"]], dim=1)
def forward_head(
self, x: list[torch.Tensor], box_head: torch.nn.Module, cls_head: torch.nn.Module, angle_head: torch.nn.Module
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and angles."""
preds = super().forward_head(x, box_head, cls_head)
if angle_head is not None:
bs = x[0].shape[0] # batch size
angle = torch.cat(
[angle_head[i](x[i]).view(bs, self.ne, -1) for i in range(self.nl)], 2
) # OBB theta logits
angle = (angle.sigmoid() - 0.25) * math.pi # [-pi/4, 3pi/4]
preds["angle"] = angle
return preds
def decode_bboxes(self, bboxes: torch.Tensor, anchors: torch.Tensor) -> torch.Tensor:
"""Decode rotated bounding boxes."""
return dist2rbox(bboxes, self.angle, anchors, dim=1)
def postprocess(self, preds: torch.Tensor) -> torch.Tensor:
"""Post-process YOLO model predictions.
Args:
preds (torch.Tensor): Raw predictions with shape (batch_size, num_anchors, 4 + nc + ne) with last dimension
format [x, y, w, h, class_probs, angle].
Returns:
(torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 7) and last
dimension format [x, y, w, h, max_class_prob, class_index, angle].
"""
boxes, scores, angle = preds.split([4, self.nc, self.ne], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
angle = angle.gather(dim=1, index=idx.repeat(1, 1, self.ne))
return torch.cat([boxes, scores, conf, angle], dim=-1)
def fuse(self) -> None:
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = self.cv4 = None
class OBB26(OBB):
"""YOLO26 OBB detection head for detection with rotation models. This class extends the OBB head with modified angle
processing that outputs raw angle predictions without sigmoid transformation, compared to the original
OBB class.
Attributes:
ne (int): Number of extra parameters.
cv4 (nn.ModuleList): Convolution layers for angle prediction.
angle (torch.Tensor): Predicted rotation angles.
Methods:
forward_head: Concatenate and return predicted bounding boxes, class probabilities, and raw angles.
Examples:
Create an OBB26 detection head
>>> obb26 = OBB26(nc=80, ne=1, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = obb26(x)
"""
def forward_head(
self, x: list[torch.Tensor], box_head: torch.nn.Module, cls_head: torch.nn.Module, angle_head: torch.nn.Module
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and raw angles."""
preds = Detect.forward_head(self, x, box_head, cls_head)
if angle_head is not None:
bs = x[0].shape[0] # batch size
angle = torch.cat(
[angle_head[i](x[i]).view(bs, self.ne, -1) for i in range(self.nl)], 2
) # OBB theta logits (raw output without sigmoid transformation)
preds["angle"] = angle
return preds
class Pose(Detect):
"""YOLO Pose head for keypoints models.
This class extends the Detect head to include keypoint prediction capabilities for pose estimation tasks.
Attributes:
kpt_shape (tuple): Number of keypoints and dimensions (2 for x,y or 3 for x,y,visible).
nk (int): Total number of keypoint values.
cv4 (nn.ModuleList): Convolution layers for keypoint prediction.
Methods:
forward: Perform forward pass through YOLO model and return predictions.
kpts_decode: Decode keypoints from predictions.
Examples:
Create a pose detection head
>>> pose = Pose(nc=80, kpt_shape=(17, 3), ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = pose(x)
"""
def __init__(self, nc: int = 80, kpt_shape: tuple = (17, 3), reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize YOLO network with default parameters and Convolutional Layers.
Args:
nc (int): Number of classes.
kpt_shape (tuple): Number of keypoints, number of dims (2 for x,y or 3 for x,y,visible).
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, reg_max, end2end, ch)
self.kpt_shape = kpt_shape # number of keypoints, number of dims (2 for x,y or 3 for x,y,visible)
self.nk = kpt_shape[0] * kpt_shape[1] # number of keypoints total
c4 = max(ch[0] // 4, self.nk)
self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, self.nk, 1)) for x in ch)
if end2end:
self.one2one_cv4 = copy.deepcopy(self.cv4)
@property
def one2many(self):
"""Returns the one-to-many head components, here for backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, pose_head=self.cv4)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(box_head=self.one2one_cv2, cls_head=self.one2one_cv3, pose_head=self.one2one_cv4)
def _inference(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Decode predicted bounding boxes and class probabilities, concatenated with keypoints."""
preds = super()._inference(x)
return torch.cat([preds, self.kpts_decode(x["kpts"])], dim=1)
def forward_head(
self, x: list[torch.Tensor], box_head: torch.nn.Module, cls_head: torch.nn.Module, pose_head: torch.nn.Module
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and keypoints."""
preds = super().forward_head(x, box_head, cls_head)
if pose_head is not None:
bs = x[0].shape[0] # batch size
preds["kpts"] = torch.cat([pose_head[i](x[i]).view(bs, self.nk, -1) for i in range(self.nl)], 2)
return preds
def postprocess(self, preds: torch.Tensor) -> torch.Tensor:
"""Post-process YOLO model predictions.
Args:
preds (torch.Tensor): Raw predictions with shape (batch_size, num_anchors, 4 + nc + nk) with last dimension
format [x1, y1, x2, y2, class_probs, keypoints].
Returns:
(torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 6 + self.nk) and
last dimension format [x1, y1, x2, y2, max_class_prob, class_index, keypoints].
"""
boxes, scores, kpts = preds.split([4, self.nc, self.nk], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
kpts = kpts.gather(dim=1, index=idx.repeat(1, 1, self.nk))
return torch.cat([boxes, scores, conf, kpts], dim=-1)
def fuse(self) -> None:
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = self.cv4 = None
def kpts_decode(self, kpts: torch.Tensor) -> torch.Tensor:
"""Decode keypoints from predictions."""
ndim = self.kpt_shape[1]
bs = kpts.shape[0]
if self.export:
y = kpts.view(bs, *self.kpt_shape, -1)
a = (y[:, :, :2] * 2.0 + (self.anchors - 0.5)) * self.strides
if ndim == 3:
a = torch.cat((a, y[:, :, 2:3].sigmoid()), 2)
return a.view(bs, self.nk, -1)
else:
y = kpts.clone()
if ndim == 3:
if NOT_MACOS14:
y[:, 2::ndim].sigmoid_()
else: # Apple macOS14 MPS bug https://github.com/ultralytics/ultralytics/pull/21878
y[:, 2::ndim] = y[:, 2::ndim].sigmoid()
y[:, 0::ndim] = (y[:, 0::ndim] * 2.0 + (self.anchors[0] - 0.5)) * self.strides
y[:, 1::ndim] = (y[:, 1::ndim] * 2.0 + (self.anchors[1] - 0.5)) * self.strides
return y
class Pose26(Pose):
"""YOLO26 Pose head for keypoints models.
This class extends the Pose head with normalizing flow for keypoint prediction in pose estimation tasks.
Attributes:
kpt_shape (tuple): Number of keypoints and dimensions (2 for x,y or 3 for x,y,visible).
nk (int): Total number of keypoint values.
cv4 (nn.ModuleList): Convolution layers for keypoint prediction.
Methods:
forward: Perform forward pass through YOLO model and return predictions.
kpts_decode: Decode keypoints from predictions.
Examples:
Create a pose detection head
>>> pose = Pose26(nc=80, kpt_shape=(17, 3), ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = pose(x)
"""
def __init__(self, nc: int = 80, kpt_shape: tuple = (17, 3), reg_max=16, end2end=False, ch: tuple = ()):
"""Initialize YOLO network with default parameters and Convolutional Layers.
Args:
nc (int): Number of classes.
kpt_shape (tuple): Number of keypoints, number of dims (2 for x,y or 3 for x,y,visible).
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, kpt_shape, reg_max, end2end, ch)
self.flow_model = RealNVP()
c4 = max(ch[0] // 4, kpt_shape[0] * (kpt_shape[1] + 2))
self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3)) for x in ch)
self.cv4_kpts = nn.ModuleList(nn.Conv2d(c4, self.nk, 1) for _ in ch)
self.nk_sigma = kpt_shape[0] * 2 # sigma_x, sigma_y for each keypoint
self.cv4_sigma = nn.ModuleList(nn.Conv2d(c4, self.nk_sigma, 1) for _ in ch)
if end2end:
self.one2one_cv4 = copy.deepcopy(self.cv4)
self.one2one_cv4_kpts = copy.deepcopy(self.cv4_kpts)
self.one2one_cv4_sigma = copy.deepcopy(self.cv4_sigma)
@property
def one2many(self):
"""Returns the one-to-many head components, here for backward compatibility."""
return dict(
box_head=self.cv2,
cls_head=self.cv3,
pose_head=self.cv4,
kpts_head=self.cv4_kpts,
kpts_sigma_head=self.cv4_sigma,
)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(
box_head=self.one2one_cv2,
cls_head=self.one2one_cv3,
pose_head=self.one2one_cv4,
kpts_head=self.one2one_cv4_kpts,
kpts_sigma_head=self.one2one_cv4_sigma,
)
def forward_head(
self,
x: list[torch.Tensor],
box_head: torch.nn.Module,
cls_head: torch.nn.Module,
pose_head: torch.nn.Module,
kpts_head: torch.nn.Module,
kpts_sigma_head: torch.nn.Module,
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and keypoints."""
preds = Detect.forward_head(self, x, box_head, cls_head)
if pose_head is not None:
bs = x[0].shape[0] # batch size
features = [pose_head[i](x[i]) for i in range(self.nl)]
preds["kpts"] = torch.cat([kpts_head[i](features[i]).view(bs, self.nk, -1) for i in range(self.nl)], 2)
if self.training:
preds["kpts_sigma"] = torch.cat(
[kpts_sigma_head[i](features[i]).view(bs, self.nk_sigma, -1) for i in range(self.nl)], 2
)
return preds
def fuse(self) -> None:
"""Remove the one2many head for inference optimization."""
super().fuse()
self.cv4_kpts = self.cv4_sigma = self.flow_model = self.one2one_cv4_sigma = None
def kpts_decode(self, kpts: torch.Tensor) -> torch.Tensor:
"""Decode keypoints from predictions."""
ndim = self.kpt_shape[1]
bs = kpts.shape[0]
if self.export:
y = kpts.view(bs, *self.kpt_shape, -1)
# NCNN fix
a = (y[:, :, :2] + self.anchors) * self.strides
if ndim == 3:
a = torch.cat((a, y[:, :, 2:3].sigmoid()), 2)
return a.view(bs, self.nk, -1)
else:
y = kpts.clone()
if ndim == 3:
if NOT_MACOS14:
y[:, 2::ndim].sigmoid_()
else: # Apple macOS14 MPS bug https://github.com/ultralytics/ultralytics/pull/21878
y[:, 2::ndim] = y[:, 2::ndim].sigmoid()
y[:, 0::ndim] = (y[:, 0::ndim] + self.anchors[0]) * self.strides
y[:, 1::ndim] = (y[:, 1::ndim] + self.anchors[1]) * self.strides
return y
class Classify(nn.Module):
"""YOLO classification head, i.e. x(b,c1,20,20) to x(b,c2).
This class implements a classification head that transforms feature maps into class predictions.
Attributes:
export (bool): Export mode flag.
conv (Conv): Convolutional layer for feature transformation.
pool (nn.AdaptiveAvgPool2d): Global average pooling layer.
drop (nn.Dropout): Dropout layer for regularization.
linear (nn.Linear): Linear layer for final classification.
Methods:
forward: Perform forward pass on input feature maps.
Examples:
Create a classification head
>>> classify = Classify(c1=1024, c2=1000)
>>> x = torch.randn(1, 1024, 20, 20)
>>> output = classify(x)
"""
export = False # export mode
def __init__(self, c1: int, c2: int, k: int = 1, s: int = 1, p: int | None = None, g: int = 1):
"""Initialize YOLO classification head to transform input tensor from (b,c1,20,20) to (b,c2) shape.
Args:
c1 (int): Number of input channels.
c2 (int): Number of output classes.
k (int): Kernel size.
s (int): Stride.
p (int, optional): Padding.
g (int): Groups.
"""
super().__init__()
c_ = 1280 # efficientnet_b0 size
self.conv = Conv(c1, c_, k, s, p, g)
self.pool = nn.AdaptiveAvgPool2d(1) # to x(b,c_,1,1)
self.drop = nn.Dropout(p=0.0, inplace=True)
self.linear = nn.Linear(c_, c2) # to x(b,c2)
def forward(self, x: list[torch.Tensor] | torch.Tensor) -> torch.Tensor | tuple:
"""Perform forward pass on input feature maps."""
if isinstance(x, list):
x = torch.cat(x, 1)
x = self.linear(self.drop(self.pool(self.conv(x)).flatten(1)))
if self.training:
return x
y = x.softmax(1) # get final output
return y if self.export else (y, x)
class WorldDetect(Detect):
"""Head for integrating YOLO detection models with semantic understanding from text embeddings.
This class extends the standard Detect head to incorporate text embeddings for enhanced semantic understanding in
object detection tasks.
Attributes:
cv3 (nn.ModuleList): Convolution layers for embedding features.
cv4 (nn.ModuleList): Contrastive head layers for text-vision alignment.
Methods:
forward: Concatenate and return predicted bounding boxes and class probabilities.
bias_init: Initialize detection head biases.
Examples:
Create a WorldDetect head
>>> world_detect = WorldDetect(nc=80, embed=512, with_bn=False, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> text = torch.randn(1, 80, 512)
>>> outputs = world_detect(x, text)
"""
def __init__(
self,
nc: int = 80,
embed: int = 512,
with_bn: bool = False,
reg_max: int = 16,
end2end: bool = False,
ch: tuple = (),
):
"""Initialize YOLO detection layer with nc classes and layer channels ch.
Args:
nc (int): Number of classes.
embed (int): Embedding dimension.
with_bn (bool): Whether to use batch normalization in contrastive head.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, reg_max=reg_max, end2end=end2end, ch=ch)
c3 = max(ch[0], min(self.nc, 100))
self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, embed, 1)) for x in ch)
self.cv4 = nn.ModuleList(BNContrastiveHead(embed) if with_bn else ContrastiveHead() for _ in ch)
def forward(self, x: list[torch.Tensor], text: torch.Tensor) -> dict[str, torch.Tensor] | tuple:
"""Concatenate and return predicted bounding boxes and class probabilities."""
feats = [xi.clone() for xi in x] # save original features for anchor generation
for i in range(self.nl):
x[i] = torch.cat((self.cv2[i](x[i]), self.cv4[i](self.cv3[i](x[i]), text)), 1)
self.no = self.nc + self.reg_max * 4 # self.nc could be changed when inference with different texts
bs = x[0].shape[0]
x_cat = torch.cat([xi.view(bs, self.no, -1) for xi in x], 2)
boxes, scores = x_cat.split((self.reg_max * 4, self.nc), 1)
preds = dict(boxes=boxes, scores=scores, feats=feats)
if self.training:
return preds
y = self._inference(preds)
return y if self.export else (y, preds)
def bias_init(self):
"""Initialize Detect() biases, WARNING: requires stride availability."""
m = self # self.model[-1] # Detect() module
# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1
# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum()) # nominal class frequency
for a, b, s in zip(m.cv2, m.cv3, m.stride): # from
a[-1].bias.data[:] = 1.0 # box
# b[-1].bias.data[:] = math.log(5 / m.nc / (640 / s) ** 2) # cls (.01 objects, 80 classes, 640 img)
class LRPCHead(nn.Module):
"""Lightweight Region Proposal and Classification Head for efficient object detection.
This head combines region proposal filtering with classification to enable efficient detection with dynamic
vocabulary support.
Attributes:
vocab (nn.Module): Vocabulary/classification layer.
pf (nn.Module): Proposal filter module.
loc (nn.Module): Localization module.
enabled (bool): Whether the head is enabled.
Methods:
conv2linear: Convert a 1x1 convolutional layer to a linear layer.
forward: Process classification and localization features to generate detection proposals.
Examples:
Create an LRPC head
>>> vocab = nn.Conv2d(256, 80, 1)
>>> pf = nn.Conv2d(256, 1, 1)
>>> loc = nn.Conv2d(256, 4, 1)
>>> head = LRPCHead(vocab, pf, loc, enabled=True)
"""
def __init__(self, vocab: nn.Module, pf: nn.Module, loc: nn.Module, enabled: bool = True):
"""Initialize LRPCHead with vocabulary, proposal filter, and localization components.
Args:
vocab (nn.Module): Vocabulary/classification module.
pf (nn.Module): Proposal filter module.
loc (nn.Module): Localization module.
enabled (bool): Whether to enable the head functionality.
"""
super().__init__()
self.vocab = self.conv2linear(vocab) if enabled else vocab
self.pf = pf
self.loc = loc
self.enabled = enabled
@staticmethod
def conv2linear(conv: nn.Conv2d) -> nn.Linear:
"""Convert a 1x1 convolutional layer to a linear layer."""
assert isinstance(conv, nn.Conv2d) and conv.kernel_size == (1, 1)
linear = nn.Linear(conv.in_channels, conv.out_channels)
linear.weight.data = conv.weight.view(conv.out_channels, -1).data
linear.bias.data = conv.bias.data
return linear
def forward(self, cls_feat: torch.Tensor, loc_feat: torch.Tensor, conf: float) -> tuple[tuple, torch.Tensor]:
"""Process classification and localization features to generate detection proposals."""
if self.enabled:
pf_score = self.pf(cls_feat)[0, 0].flatten(0)
mask = pf_score.sigmoid() > conf
cls_feat = cls_feat.flatten(2).transpose(-1, -2)
cls_feat = self.vocab(cls_feat[:, mask] if conf else cls_feat * mask.unsqueeze(-1).int())
return self.loc(loc_feat), cls_feat.transpose(-1, -2), mask
else:
cls_feat = self.vocab(cls_feat)
loc_feat = self.loc(loc_feat)
return (
loc_feat,
cls_feat.flatten(2),
torch.ones(cls_feat.shape[2] * cls_feat.shape[3], device=cls_feat.device, dtype=torch.bool),
)
class YOLOEDetect(Detect):
"""Head for integrating YOLO detection models with semantic understanding from text embeddings.
This class extends the standard Detect head to support text-guided detection with enhanced semantic understanding
through text embeddings and visual prompt embeddings.
Attributes:
is_fused (bool): Whether the model is fused for inference.
cv3 (nn.ModuleList): Convolution layers for embedding features.
cv4 (nn.ModuleList): Contrastive head layers for text-vision alignment.
reprta (Residual): Residual block for text prompt embeddings.
savpe (SAVPE): Spatial-aware visual prompt embeddings module.
embed (int): Embedding dimension.
Methods:
fuse: Fuse text features with model weights for efficient inference.
get_tpe: Get text prompt embeddings with normalization.
get_vpe: Get visual prompt embeddings with spatial awareness.
forward_lrpc: Process features with fused text embeddings for prompt-free model.
forward: Process features with class prompt embeddings to generate detections.
bias_init: Initialize biases for detection heads.
Examples:
Create a YOLOEDetect head
>>> yoloe_detect = YOLOEDetect(nc=80, embed=512, with_bn=True, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> cls_pe = torch.randn(1, 80, 512)
>>> outputs = yoloe_detect(x, cls_pe)
"""
is_fused = False
def __init__(
self, nc: int = 80, embed: int = 512, with_bn: bool = False, reg_max=16, end2end=False, ch: tuple = ()
):
"""Initialize YOLO detection layer with nc classes and layer channels ch.
Args:
nc (int): Number of classes.
embed (int): Embedding dimension.
with_bn (bool): Whether to use batch normalization in contrastive head.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, reg_max, end2end, ch)
c3 = max(ch[0], min(self.nc, 100))
assert c3 <= embed
assert with_bn
self.cv3 = (
nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, embed, 1)) for x in ch)
if self.legacy
else nn.ModuleList(
nn.Sequential(
nn.Sequential(DWConv(x, x, 3), Conv(x, c3, 1)),
nn.Sequential(DWConv(c3, c3, 3), Conv(c3, c3, 1)),
nn.Conv2d(c3, embed, 1),
)
for x in ch
)
)
self.cv4 = nn.ModuleList(BNContrastiveHead(embed) if with_bn else ContrastiveHead() for _ in ch)
if end2end:
self.one2one_cv3 = copy.deepcopy(self.cv3) # overwrite with new cv3
self.one2one_cv4 = copy.deepcopy(self.cv4)
self.reprta = Residual(SwiGLUFFN(embed, embed))
self.savpe = SAVPE(ch, c3, embed)
self.embed = embed
@smart_inference_mode()
def fuse(self, txt_feats: torch.Tensor = None):
"""Fuse text features with model weights for efficient inference."""
if txt_feats is None: # means eliminate one2many branch
self.cv2 = self.cv3 = self.cv4 = None
return
if self.is_fused:
return
assert not self.training
txt_feats = txt_feats.to(torch.float32).squeeze(0)
self._fuse_tp(txt_feats, self.cv3, self.cv4)
if self.end2end:
self._fuse_tp(txt_feats, self.one2one_cv3, self.one2one_cv4)
del self.reprta
self.reprta = nn.Identity()
self.is_fused = True
def _fuse_tp(self, txt_feats: torch.Tensor, cls_head: torch.nn.Module, bn_head: torch.nn.Module) -> None:
"""Fuse text prompt embeddings with model weights for efficient inference."""
for cls_h, bn_h in zip(cls_head, bn_head):
assert isinstance(cls_h, nn.Sequential)
assert isinstance(bn_h, BNContrastiveHead)
conv = cls_h[-1]
assert isinstance(conv, nn.Conv2d)
logit_scale = bn_h.logit_scale
bias = bn_h.bias
norm = bn_h.norm
t = txt_feats * logit_scale.exp()
conv: nn.Conv2d = fuse_conv_and_bn(conv, norm)
w = conv.weight.data.squeeze(-1).squeeze(-1)
b = conv.bias.data
w = t @ w
b1 = (t @ b.reshape(-1).unsqueeze(-1)).squeeze(-1)
b2 = torch.ones_like(b1) * bias
conv = (
nn.Conv2d(
conv.in_channels,
w.shape[0],
kernel_size=1,
)
.requires_grad_(False)
.to(conv.weight.device)
)
conv.weight.data.copy_(w.unsqueeze(-1).unsqueeze(-1))
conv.bias.data.copy_(b1 + b2)
cls_h[-1] = conv
bn_h.fuse()
def get_tpe(self, tpe: torch.Tensor | None) -> torch.Tensor | None:
"""Get text prompt embeddings with normalization."""
return None if tpe is None else F.normalize(self.reprta(tpe), dim=-1, p=2)
def get_vpe(self, x: list[torch.Tensor], vpe: torch.Tensor) -> torch.Tensor:
"""Get visual prompt embeddings with spatial awareness."""
if vpe.shape[1] == 0: # no visual prompt embeddings
return torch.zeros(x[0].shape[0], 0, self.embed, device=x[0].device)
if vpe.ndim == 4: # (B, N, H, W)
vpe = self.savpe(x, vpe)
assert vpe.ndim == 3 # (B, N, D)
return vpe
def forward(self, x: list[torch.Tensor]) -> torch.Tensor | tuple:
"""Process features with class prompt embeddings to generate detections."""
if hasattr(self, "lrpc"): # for prompt-free inference
return self.forward_lrpc(x[:3])
return super().forward(x)
def forward_lrpc(self, x: list[torch.Tensor]) -> torch.Tensor | tuple:
"""Process features with fused text embeddings to generate detections for prompt-free model."""
boxes, scores, index = [], [], []
bs = x[0].shape[0]
cv2 = self.cv2 if not self.end2end else self.one2one_cv2
cv3 = self.cv3 if not self.end2end else self.one2one_cv3
for i in range(self.nl):
cls_feat = cv3[i](x[i])
loc_feat = cv2[i](x[i])
assert isinstance(self.lrpc[i], LRPCHead)
box, score, idx = self.lrpc[i](
cls_feat,
loc_feat,
0 if self.export and not self.dynamic else getattr(self, "conf", 0.001),
)
boxes.append(box.view(bs, self.reg_max * 4, -1))
scores.append(score)
index.append(idx)
preds = dict(boxes=torch.cat(boxes, 2), scores=torch.cat(scores, 2), feats=x, index=torch.cat(index))
y = self._inference(preds)
if self.end2end:
y = self.postprocess(y.permute(0, 2, 1))
return y if self.export else (y, preds)
def _get_decode_boxes(self, x):
"""Decode predicted bounding boxes for inference."""
dbox = super()._get_decode_boxes(x)
if hasattr(self, "lrpc"):
dbox = dbox if self.export and not self.dynamic else dbox[..., x["index"]]
return dbox
@property
def one2many(self):
"""Returns the one-to-many head components, here for v3/v5/v8/v9/v11 backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, contrastive_head=self.cv4)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(box_head=self.one2one_cv2, cls_head=self.one2one_cv3, contrastive_head=self.one2one_cv4)
def forward_head(self, x, box_head, cls_head, contrastive_head):
"""Concatenates and returns predicted bounding boxes, class probabilities, and contrastive scores."""
assert len(x) == 4, f"Expected 4 features including 3 feature maps and 1 text embeddings, but got {len(x)}."
if box_head is None or cls_head is None: # for fused inference
return dict()
bs = x[0].shape[0] # batch size
boxes = torch.cat([box_head[i](x[i]).view(bs, 4 * self.reg_max, -1) for i in range(self.nl)], dim=-1)
self.nc = x[-1].shape[1]
scores = torch.cat(
[contrastive_head[i](cls_head[i](x[i]), x[-1]).reshape(bs, self.nc, -1) for i in range(self.nl)], dim=-1
)
self.no = self.nc + self.reg_max * 4 # self.nc could be changed when inference with different texts
return dict(boxes=boxes, scores=scores, feats=x[:3])
def bias_init(self):
"""Initialize Detect() biases, WARNING: requires stride availability."""
for i, (a, b, c) in enumerate(
zip(self.one2many["box_head"], self.one2many["cls_head"], self.one2many["contrastive_head"])
):
a[-1].bias.data[:] = 2.0 # box
b[-1].bias.data[:] = 0.0
c.bias.data[:] = math.log(5 / self.nc / (640 / self.stride[i]) ** 2)
if self.end2end:
for i, (a, b, c) in enumerate(
zip(self.one2one["box_head"], self.one2one["cls_head"], self.one2one["contrastive_head"])
):
a[-1].bias.data[:] = 2.0 # box
b[-1].bias.data[:] = 0.0
c.bias.data[:] = math.log(5 / self.nc / (640 / self.stride[i]) ** 2)
class YOLOESegment(YOLOEDetect):
"""YOLO segmentation head with text embedding capabilities.
This class extends YOLOEDetect to include mask prediction capabilities for instance segmentation tasks with
text-guided semantic understanding.
Attributes:
nm (int): Number of masks.
npr (int): Number of protos.
proto (Proto): Prototype generation module.
cv5 (nn.ModuleList): Convolution layers for mask coefficients.
Methods:
forward: Return model outputs and mask coefficients.
Examples:
Create a YOLOESegment head
>>> yoloe_segment = YOLOESegment(nc=80, nm=32, npr=256, embed=512, with_bn=True, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> text = torch.randn(1, 80, 512)
>>> outputs = yoloe_segment(x, text)
"""
def __init__(
self,
nc: int = 80,
nm: int = 32,
npr: int = 256,
embed: int = 512,
with_bn: bool = False,
reg_max=16,
end2end=False,
ch: tuple = (),
):
"""Initialize YOLOESegment with class count, mask parameters, and embedding dimensions.
Args:
nc (int): Number of classes.
nm (int): Number of masks.
npr (int): Number of protos.
embed (int): Embedding dimension.
with_bn (bool): Whether to use batch normalization in contrastive head.
reg_max (int): Maximum number of DFL channels.
end2end (bool): Whether to use end-to-end NMS-free detection.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, embed, with_bn, reg_max, end2end, ch)
self.nm = nm
self.npr = npr
self.proto = Proto(ch[0], self.npr, self.nm)
c5 = max(ch[0] // 4, self.nm)
self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nm, 1)) for x in ch)
if end2end:
self.one2one_cv5 = copy.deepcopy(self.cv5)
@property
def one2many(self):
"""Returns the one-to-many head components, here for v3/v5/v8/v9/v11 backward compatibility."""
return dict(box_head=self.cv2, cls_head=self.cv3, mask_head=self.cv5, contrastive_head=self.cv4)
@property
def one2one(self):
"""Returns the one-to-one head components."""
return dict(
box_head=self.one2one_cv2,
cls_head=self.one2one_cv3,
mask_head=self.one2one_cv5,
contrastive_head=self.one2one_cv4,
)
def forward_lrpc(self, x: list[torch.Tensor]) -> torch.Tensor | tuple:
"""Process features with fused text embeddings to generate detections for prompt-free model."""
boxes, scores, index = [], [], []
bs = x[0].shape[0]
cv2 = self.cv2 if not self.end2end else self.one2one_cv2
cv3 = self.cv3 if not self.end2end else self.one2one_cv3
cv5 = self.cv5 if not self.end2end else self.one2one_cv5
for i in range(self.nl):
cls_feat = cv3[i](x[i])
loc_feat = cv2[i](x[i])
assert isinstance(self.lrpc[i], LRPCHead)
box, score, idx = self.lrpc[i](
cls_feat,
loc_feat,
0 if self.export and not self.dynamic else getattr(self, "conf", 0.001),
)
boxes.append(box.view(bs, self.reg_max * 4, -1))
scores.append(score)
index.append(idx)
mc = torch.cat([cv5[i](x[i]).view(bs, self.nm, -1) for i in range(self.nl)], 2)
index = torch.cat(index)
preds = dict(
boxes=torch.cat(boxes, 2),
scores=torch.cat(scores, 2),
feats=x,
index=index,
mask_coefficient=mc * index.int() if self.export and not self.dynamic else mc[..., index],
)
y = self._inference(preds)
if self.end2end:
y = self.postprocess(y.permute(0, 2, 1))
return y if self.export else (y, preds)
def forward(self, x: list[torch.Tensor]) -> tuple | list[torch.Tensor] | dict[str, torch.Tensor]:
"""Return model outputs and mask coefficients if training, otherwise return outputs and mask coefficients."""
outputs = super().forward(x)
preds = outputs[1] if isinstance(outputs, tuple) else outputs
proto = self.proto(x[0]) # mask protos
if isinstance(preds, dict): # training and validating during training
if self.end2end:
preds["one2many"]["proto"] = proto
preds["one2one"]["proto"] = proto.detach()
else:
preds["proto"] = proto
if self.training:
return preds
return (outputs, proto) if self.export else ((outputs[0], proto), preds)
def _inference(self, x: dict[str, torch.Tensor]) -> torch.Tensor:
"""Decode predicted bounding boxes and class probabilities, concatenated with mask coefficients."""
preds = super()._inference(x)
return torch.cat([preds, x["mask_coefficient"]], dim=1)
def forward_head(
self,
x: list[torch.Tensor],
box_head: torch.nn.Module,
cls_head: torch.nn.Module,
mask_head: torch.nn.Module,
contrastive_head: torch.nn.Module,
) -> dict[str, torch.Tensor]:
"""Concatenates and returns predicted bounding boxes, class probabilities, and mask coefficients."""
preds = super().forward_head(x, box_head, cls_head, contrastive_head)
if mask_head is not None:
bs = x[0].shape[0] # batch size
preds["mask_coefficient"] = torch.cat([mask_head[i](x[i]).view(bs, self.nm, -1) for i in range(self.nl)], 2)
return preds
def postprocess(self, preds: torch.Tensor) -> torch.Tensor:
"""Post-process YOLO model predictions.
Args:
preds (torch.Tensor): Raw predictions with shape (batch_size, num_anchors, 4 + nc + nm) with last dimension
format [x1, y1, x2, y2, class_probs, mask_coefficient].
Returns:
(torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 6 + nm) and last
dimension format [x1, y1, x2, y2, max_class_prob, class_index, mask_coefficient].
"""
boxes, scores, mask_coefficient = preds.split([4, self.nc, self.nm], dim=-1)
scores, conf, idx = self.get_topk_index(scores, self.max_det)
boxes = boxes.gather(dim=1, index=idx.repeat(1, 1, 4))
mask_coefficient = mask_coefficient.gather(dim=1, index=idx.repeat(1, 1, self.nm))
return torch.cat([boxes, scores, conf, mask_coefficient], dim=-1)
def fuse(self, txt_feats: torch.Tensor = None):
"""Fuse text features with model weights for efficient inference."""
super().fuse(txt_feats)
if txt_feats is None: # means eliminate one2many branch
self.cv5 = None
if hasattr(self.proto, "fuse"):
self.proto.fuse()
return
class YOLOESegment26(YOLOESegment):
"""YOLOE-style segmentation head module using Proto26 for mask generation.
This class extends the YOLOESegment functionality to include segmentation capabilities by integrating a Proto26
generation module and convolutional layers to predict mask coefficients.
Args:
nc (int): Number of classes. Defaults to 80.
nm (int): Number of masks. Defaults to 32.
npr (int): Number of prototype channels. Defaults to 256.
embed (int): Embedding dimensionality. Defaults to 512.
with_bn (bool): Whether to use Batch Normalization. Defaults to False.
reg_max (int): Maximum number of DFL channels. Defaults to 16.
end2end (bool): Whether to use end-to-end detection mode. Defaults to False.
ch (tuple[int, ...]): Input channels for each scale.
Attributes:
nm (int): Number of segmentation masks.
npr (int): Number of prototype channels.
proto (Proto26): Prototype generation module for segmentation.
cv5 (nn.ModuleList): Convolutional layers for generating mask coefficients from features.
one2one_cv5 (nn.ModuleList, optional): Deep copy of cv5 for end-to-end detection branches.
"""
def __init__(
self,
nc: int = 80,
nm: int = 32,
npr: int = 256,
embed: int = 512,
with_bn: bool = False,
reg_max=16,
end2end=False,
ch: tuple = (),
):
"""Initialize YOLOESegment26 with class count, mask parameters, and embedding dimensions."""
YOLOEDetect.__init__(self, nc, embed, with_bn, reg_max, end2end, ch)
self.nm = nm
self.npr = npr
self.proto = Proto26(ch, self.npr, self.nm, nc) # protos
c5 = max(ch[0] // 4, self.nm)
self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nm, 1)) for x in ch)
if end2end:
self.one2one_cv5 = copy.deepcopy(self.cv5)
def forward(self, x: list[torch.Tensor]) -> tuple | list[torch.Tensor] | dict[str, torch.Tensor]:
"""Return model outputs and mask coefficients if training, otherwise return outputs and mask coefficients."""
outputs = YOLOEDetect.forward(self, x)
preds = outputs[1] if isinstance(outputs, tuple) else outputs
proto = self.proto([xi.detach() for xi in x], return_semseg=False) # mask protos
if isinstance(preds, dict): # training and validating during training
if self.end2end and not hasattr(self, "lrpc"): # not prompt-free
preds["one2many"]["proto"] = proto
preds["one2one"]["proto"] = proto.detach()
else:
preds["proto"] = proto
if self.training:
return preds
return (outputs, proto) if self.export else ((outputs[0], proto), preds)
class RTDETRDecoder(nn.Module):
"""Real-Time Deformable Transformer Decoder (RTDETRDecoder) module for object detection.
This decoder module utilizes Transformer architecture along with deformable convolutions to predict bounding boxes
and class labels for objects in an image. It integrates features from multiple layers and runs through a series of
Transformer decoder layers to output the final predictions.
Attributes:
export (bool): Export mode flag.
hidden_dim (int): Dimension of hidden layers.
nhead (int): Number of heads in multi-head attention.
nl (int): Number of feature levels.
nc (int): Number of classes.
num_queries (int): Number of query points.
num_decoder_layers (int): Number of decoder layers.
input_proj (nn.ModuleList): Input projection layers for backbone features.
decoder (DeformableTransformerDecoder): Transformer decoder module.
denoising_class_embed (nn.Embedding): Class embeddings for denoising.
num_denoising (int): Number of denoising queries.
label_noise_ratio (float): Label noise ratio for training.
box_noise_scale (float): Box noise scale for training.
learnt_init_query (bool): Whether to learn initial query embeddings.
tgt_embed (nn.Embedding): Target embeddings for queries.
query_pos_head (MLP): Query position head.
enc_output (nn.Sequential): Encoder output layers.
enc_score_head (nn.Linear): Encoder score prediction head.
enc_bbox_head (MLP): Encoder bbox prediction head.
dec_score_head (nn.ModuleList): Decoder score prediction heads.
dec_bbox_head (nn.ModuleList): Decoder bbox prediction heads.
Methods:
forward: Run forward pass and return bounding box and classification scores.
Examples:
Create an RTDETRDecoder
>>> decoder = RTDETRDecoder(nc=80, ch=(512, 1024, 2048), hd=256, nq=300)
>>> x = [torch.randn(1, 512, 64, 64), torch.randn(1, 1024, 32, 32), torch.randn(1, 2048, 16, 16)]
>>> outputs = decoder(x)
"""
export = False # export mode
shapes = []
anchors = torch.empty(0)
valid_mask = torch.empty(0)
dynamic = False
def __init__(
self,
nc: int = 80,
ch: tuple = (512, 1024, 2048),
hd: int = 256, # hidden dim
nq: int = 300, # num queries
ndp: int = 4, # num decoder points
nh: int = 8, # num head
ndl: int = 6, # num decoder layers
d_ffn: int = 1024, # dim of feedforward
dropout: float = 0.0,
act: nn.Module = nn.ReLU(),
eval_idx: int = -1,
# Training args
nd: int = 100, # num denoising
label_noise_ratio: float = 0.5,
box_noise_scale: float = 1.0,
learnt_init_query: bool = False,
):
"""Initialize the RTDETRDecoder module with the given parameters.
Args:
nc (int): Number of classes.
ch (tuple): Channels in the backbone feature maps.
hd (int): Dimension of hidden layers.
nq (int): Number of query points.
ndp (int): Number of decoder points.
nh (int): Number of heads in multi-head attention.
ndl (int): Number of decoder layers.
d_ffn (int): Dimension of the feed-forward networks.
dropout (float): Dropout rate.
act (nn.Module): Activation function.
eval_idx (int): Evaluation index.
nd (int): Number of denoising.
label_noise_ratio (float): Label noise ratio.
box_noise_scale (float): Box noise scale.
learnt_init_query (bool): Whether to learn initial query embeddings.
"""
super().__init__()
self.hidden_dim = hd
self.nhead = nh
self.nl = len(ch) # num level
self.nc = nc
self.num_queries = nq
self.num_decoder_layers = ndl
# Backbone feature projection
self.input_proj = nn.ModuleList(nn.Sequential(nn.Conv2d(x, hd, 1, bias=False), nn.BatchNorm2d(hd)) for x in ch)
# NOTE: simplified version but it's not consistent with .pt weights.
# self.input_proj = nn.ModuleList(Conv(x, hd, act=False) for x in ch)
# Transformer module
decoder_layer = DeformableTransformerDecoderLayer(hd, nh, d_ffn, dropout, act, self.nl, ndp)
self.decoder = DeformableTransformerDecoder(hd, decoder_layer, ndl, eval_idx)
# Denoising part
self.denoising_class_embed = nn.Embedding(nc, hd)
self.num_denoising = nd
self.label_noise_ratio = label_noise_ratio
self.box_noise_scale = box_noise_scale
# Decoder embedding
self.learnt_init_query = learnt_init_query
if learnt_init_query:
self.tgt_embed = nn.Embedding(nq, hd)
self.query_pos_head = MLP(4, 2 * hd, hd, num_layers=2)
# Encoder head
self.enc_output = nn.Sequential(nn.Linear(hd, hd), nn.LayerNorm(hd))
self.enc_score_head = nn.Linear(hd, nc)
self.enc_bbox_head = MLP(hd, hd, 4, num_layers=3)
# Decoder head
self.dec_score_head = nn.ModuleList([nn.Linear(hd, nc) for _ in range(ndl)])
self.dec_bbox_head = nn.ModuleList([MLP(hd, hd, 4, num_layers=3) for _ in range(ndl)])
self._reset_parameters()
def forward(self, x: list[torch.Tensor], batch: dict | None = None) -> tuple | torch.Tensor:
"""Run the forward pass of the module, returning bounding box and classification scores for the input.
Args:
x (list[torch.Tensor]): List of feature maps from the backbone.
batch (dict, optional): Batch information for training.
Returns:
outputs (tuple | torch.Tensor): During training, returns a tuple of bounding boxes, scores, and other
metadata. During inference, returns a tensor of shape (bs, 300, 4+nc) containing bounding boxes and
class scores.
"""
from ultralytics.models.utils.ops import get_cdn_group
# Input projection and embedding
feats, shapes = self._get_encoder_input(x)
# Prepare denoising training
dn_embed, dn_bbox, attn_mask, dn_meta = get_cdn_group(
batch,
self.nc,
self.num_queries,
self.denoising_class_embed.weight,
self.num_denoising,
self.label_noise_ratio,
self.box_noise_scale,
self.training,
)
embed, refer_bbox, enc_bboxes, enc_scores = self._get_decoder_input(feats, shapes, dn_embed, dn_bbox)
# Decoder
dec_bboxes, dec_scores = self.decoder(
embed,
refer_bbox,
feats,
shapes,
self.dec_bbox_head,
self.dec_score_head,
self.query_pos_head,
attn_mask=attn_mask,
)
x = dec_bboxes, dec_scores, enc_bboxes, enc_scores, dn_meta
if self.training:
return x
# (bs, 300, 4+nc)
y = torch.cat((dec_bboxes.squeeze(0), dec_scores.squeeze(0).sigmoid()), -1)
return y if self.export else (y, x)
@staticmethod
def _generate_anchors(
shapes: list[list[int]],
grid_size: float = 0.05,
dtype: torch.dtype = torch.float32,
device: str = "cpu",
eps: float = 1e-2,
) -> tuple[torch.Tensor, torch.Tensor]:
"""Generate anchor bounding boxes for given shapes with specific grid size and validate them.
Args:
shapes (list): List of feature map shapes.
grid_size (float, optional): Base size of grid cells.
dtype (torch.dtype, optional): Data type for tensors.
device (str, optional): Device to create tensors on.
eps (float, optional): Small value for numerical stability.
Returns:
anchors (torch.Tensor): Generated anchor boxes.
valid_mask (torch.Tensor): Valid mask for anchors.
"""
anchors = []
for i, (h, w) in enumerate(shapes):
sy = torch.arange(end=h, dtype=dtype, device=device)
sx = torch.arange(end=w, dtype=dtype, device=device)
grid_y, grid_x = torch.meshgrid(sy, sx, indexing="ij") if TORCH_1_11 else torch.meshgrid(sy, sx)
grid_xy = torch.stack([grid_x, grid_y], -1) # (h, w, 2)
valid_WH = torch.tensor([w, h], dtype=dtype, device=device)
grid_xy = (grid_xy.unsqueeze(0) + 0.5) / valid_WH # (1, h, w, 2)
wh = torch.ones_like(grid_xy, dtype=dtype, device=device) * grid_size * (2.0**i)
anchors.append(torch.cat([grid_xy, wh], -1).view(-1, h * w, 4)) # (1, h*w, 4)
anchors = torch.cat(anchors, 1) # (1, h*w*nl, 4)
valid_mask = ((anchors > eps) & (anchors < 1 - eps)).all(-1, keepdim=True) # 1, h*w*nl, 1
anchors = torch.log(anchors / (1 - anchors))
anchors = anchors.masked_fill(~valid_mask, float("inf"))
return anchors, valid_mask
def _get_encoder_input(self, x: list[torch.Tensor]) -> tuple[torch.Tensor, list[list[int]]]:
"""Process and return encoder inputs by getting projection features from input and concatenating them.
Args:
x (list[torch.Tensor]): List of feature maps from the backbone.
Returns:
feats (torch.Tensor): Processed features.
shapes (list): List of feature map shapes.
"""
# Get projection features
x = [self.input_proj[i](feat) for i, feat in enumerate(x)]
# Get encoder inputs
feats = []
shapes = []
for feat in x:
h, w = feat.shape[2:]
# [b, c, h, w] -> [b, h*w, c]
feats.append(feat.flatten(2).permute(0, 2, 1))
# [nl, 2]
shapes.append([h, w])
# [b, h*w, c]
feats = torch.cat(feats, 1)
return feats, shapes
def _get_decoder_input(
self,
feats: torch.Tensor,
shapes: list[list[int]],
dn_embed: torch.Tensor | None = None,
dn_bbox: torch.Tensor | None = None,
) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
"""Generate and prepare the input required for the decoder from the provided features and shapes.
Args:
feats (torch.Tensor): Processed features from encoder.
shapes (list): List of feature map shapes.
dn_embed (torch.Tensor, optional): Denoising embeddings.
dn_bbox (torch.Tensor, optional): Denoising bounding boxes.
Returns:
embeddings (torch.Tensor): Query embeddings for decoder.
refer_bbox (torch.Tensor): Reference bounding boxes.
enc_bboxes (torch.Tensor): Encoded bounding boxes.
enc_scores (torch.Tensor): Encoded scores.
"""
bs = feats.shape[0]
if self.dynamic or self.shapes != shapes:
self.anchors, self.valid_mask = self._generate_anchors(shapes, dtype=feats.dtype, device=feats.device)
self.shapes = shapes
# Prepare input for decoder
features = self.enc_output(self.valid_mask * feats) # bs, h*w, 256
enc_outputs_scores = self.enc_score_head(features) # (bs, h*w, nc)
# Query selection
# (bs*num_queries,)
topk_ind = torch.topk(enc_outputs_scores.max(-1).values, self.num_queries, dim=1).indices.view(-1)
# (bs*num_queries,)
batch_ind = torch.arange(end=bs, dtype=topk_ind.dtype).unsqueeze(-1).repeat(1, self.num_queries).view(-1)
# (bs, num_queries, 256)
top_k_features = features[batch_ind, topk_ind].view(bs, self.num_queries, -1)
# (bs, num_queries, 4)
top_k_anchors = self.anchors[:, topk_ind].view(bs, self.num_queries, -1)
# Dynamic anchors + static content
refer_bbox = self.enc_bbox_head(top_k_features) + top_k_anchors
enc_bboxes = refer_bbox.sigmoid()
if dn_bbox is not None:
refer_bbox = torch.cat([dn_bbox, refer_bbox], 1)
enc_scores = enc_outputs_scores[batch_ind, topk_ind].view(bs, self.num_queries, -1)
embeddings = self.tgt_embed.weight.unsqueeze(0).repeat(bs, 1, 1) if self.learnt_init_query else top_k_features
if self.training:
refer_bbox = refer_bbox.detach()
if not self.learnt_init_query:
embeddings = embeddings.detach()
if dn_embed is not None:
embeddings = torch.cat([dn_embed, embeddings], 1)
return embeddings, refer_bbox, enc_bboxes, enc_scores
def _reset_parameters(self):
"""Initialize or reset the parameters of the model's various components with predefined weights and biases."""
# Class and bbox head init
bias_cls = bias_init_with_prob(0.01) / 80 * self.nc
# NOTE: the weight initialization in `linear_init` would cause NaN when training with custom datasets.
# linear_init(self.enc_score_head)
constant_(self.enc_score_head.bias, bias_cls)
constant_(self.enc_bbox_head.layers[-1].weight, 0.0)
constant_(self.enc_bbox_head.layers[-1].bias, 0.0)
for cls_, reg_ in zip(self.dec_score_head, self.dec_bbox_head):
# linear_init(cls_)
constant_(cls_.bias, bias_cls)
constant_(reg_.layers[-1].weight, 0.0)
constant_(reg_.layers[-1].bias, 0.0)
linear_init(self.enc_output[0])
xavier_uniform_(self.enc_output[0].weight)
if self.learnt_init_query:
xavier_uniform_(self.tgt_embed.weight)
xavier_uniform_(self.query_pos_head.layers[0].weight)
xavier_uniform_(self.query_pos_head.layers[1].weight)
for layer in self.input_proj:
xavier_uniform_(layer[0].weight)
class v10Detect(Detect):
"""v10 Detection head from https://arxiv.org/pdf/2405.14458.
This class implements the YOLOv10 detection head with dual-assignment training and consistent dual predictions for
improved efficiency and performance.
Attributes:
end2end (bool): End-to-end detection mode.
max_det (int): Maximum number of detections.
cv3 (nn.ModuleList): Light classification head layers.
one2one_cv3 (nn.ModuleList): One-to-one classification head layers.
Methods:
__init__: Initialize the v10Detect object with specified number of classes and input channels.
forward: Perform forward pass of the v10Detect module.
bias_init: Initialize biases of the Detect module.
fuse: Remove the one2many head for inference optimization.
Examples:
Create a v10Detect head
>>> v10_detect = v10Detect(nc=80, ch=(256, 512, 1024))
>>> x = [torch.randn(1, 256, 80, 80), torch.randn(1, 512, 40, 40), torch.randn(1, 1024, 20, 20)]
>>> outputs = v10_detect(x)
"""
end2end = True
def __init__(self, nc: int = 80, ch: tuple = ()):
"""Initialize the v10Detect object with the specified number of classes and input channels.
Args:
nc (int): Number of classes.
ch (tuple): Tuple of channel sizes from backbone feature maps.
"""
super().__init__(nc, end2end=True, ch=ch)
c3 = max(ch[0], min(self.nc, 100)) # channels
# Light cls head
self.cv3 = nn.ModuleList(
nn.Sequential(
nn.Sequential(Conv(x, x, 3, g=x), Conv(x, c3, 1)),
nn.Sequential(Conv(c3, c3, 3, g=c3), Conv(c3, c3, 1)),
nn.Conv2d(c3, self.nc, 1),
)
for x in ch
)
self.one2one_cv3 = copy.deepcopy(self.cv3)
def fuse(self):
"""Remove the one2many head for inference optimization."""
self.cv2 = self.cv3 = None