# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

# Ground 2D Detection Dataset for Mono3D
# Custom annotation format with difficulty scores and class mapping

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: /mnt/nfs/mono3d/ydong_data/Detection/Detection2D_20260427  # dataset root dir
train: /mnt/nfs/mono3d/ydong_data/Detection/Detection2D_20260427/train.txt  # train images
val: /mnt/nfs/mono3d/ydong_data/Detection/Detection2D_20260427/val.txt  # val images
test:  # test images (optional)

# Class mapping: string class names to numeric IDs
# Format: class_name: class_id (allows easy merging, e.g., car: 0, van: 0)
class_map:
  car: 0
  suv: 1
  pickup: 2
  medium_car: 3
  van: 4
  bus: 5    
  truck: 6
  tanker: 6
  large_truck: 6
  construction_vehicle: 6
  special_vehicle: 7
  unknown: 8
  pedestrian: 9
  bicyclist: 10
  motorcyclist: 10
  bicycle: 11
  motorcycle: 11
  tricycle: 12
  tricyclist: 12
  traffic_sign: 13
  wheel: 14
  plate: 15
  face: 16
  car_fake: 17
  bicyclist_fake: 18
  pedestrian_fake: 19
  car_carrier: 6
  platform_truck: 6


# Training parameters
min_wh: 8.0  # Keep boxes whose width or height is at least this many pixels
             # Recommended: 2 * smallest_stride (2 * 8 = 16) for network detectability

# Color space of input images
use_yuv444: false  # Convert YUV444 to BGR in dataloader (BT.601 full range)

# Label file format (7 columns):
# [class_name x_center y_center width height difficulty1 difficulty2]

# Difficulty-based loss weighting:      difficulty_weights 设计上是给 0/1/2/3 难度目标配置权重的，但当前 Ground 2D 检测的 box/cls/dfl loss 没有实际按它加权；当前 difficulty 主要作为额外 difficulty 二分类监督使用
difficulty_weights: [1.0, 1.0, 0.7, 0.3]