Prepare Your Object Detector

  1. Install
  2. Annotate
  3. Config and Train YOLOv5
  4. Extract Objects from Images
  5. Train Your Human+Hand+Face Detector

In this tutorial, we will show how to annotate and train an object detector for your custom objects.


mkdir -p 3rdparty && cd 3rdparty
git clone
mv models/


Create bbox by drag and press n.

python3 scripts/postprocess/ ${data} /home/qing/DGPU/nas/datasets/soccer/objects --annot ball

Config and Train YOLOv5

python3 --img 640 --batch 16 --epochs 200 --data data/anyobject.yml --weights models/ --name anyobject

Extract Objects from Images

python3 apps/preprocess/ ${data} --ckpt 3rdparty/yolov5/runs/train/anyobject/weights/ --out object2d

Train Your Human+Hand+Face Detector