YoloV9:2D bounding boxes and keypoints detection

Create Dataset

1. X-Anylabeling(Version 2.5.4)
2. Installation tutorial: Bilibili
3. label

   [class_ID x_center y_center bbox_width bbox_height x1 y1 vis x2 y2 vis ...]
   

Important files in the Yolov9 pose detection algorithm file

1. datasets/pose_json2txt.py: Transfer label files from .json to .txt
2. datasets/split.py: split dataset
3. change configuration files

• ultralytics\cfg\models\v9\yolov9-pose.yaml

  # Parameters
  nc: 1  # number of classes
  kpt_shape: [2, 3] # number of keypoints, just change the first parameter, '3' means visibility
  

• ultralytics\cfg\datasets\coco-pose.yaml

  # Keypoints
  kpt_shape: [2, 3] # number of keypoints, number of dims (2 for x,y or 3 for x,y,visible)
  # Classes
  names:
  0: Joint
  

• ultralytics\utils\plotting.py

  # line 243 'radius' to control the size of keypoints
  cv2.circle(self.im, (int(x_coord), int(y_coord)), radius, color_k, -1, lineType=cv2.LINE_AA)
  # cv2.circle(self.im, (int(x_coord), int(y_coord)), , color_k, -1, lineType=cv2.LINE_AA)
  

• ultralytics\cfg\default.yaml: default configuration file

Train

yolo pose train data=ultralytics\cfg\datasets\coco-pose.yaml model=ultralytics\cfg\models\v9\yolov9-pose.yaml epochs=300 imgsz=640

---

Related links

code:Github / Our code
dataset: