Contents

- DINO Version3 Test Sample Code

• Download Weights

How to Run

• Run env

conda env create -f conda.yaml
conda activate dinov3
pip install -r requirements.txt

• Run Feature Extraction
  • Build '.env' file for Hugging face token

python demo.py

• Run Detection
  • Weights of backbone path: 'weights/dinov3_vit7b16_pretrain_lvd1689m-a955f4ea.pth'
  • Weights of detector head path: 'weights/dinov3_vit7b16_coco_detr_head-b0235ff7.pth'
  • Adjust fps_in variables

python demo_detection.py --video_path {input_video_path} --conf_thresh {conf_score_thr} --max_time_sec {max_seconds_of_ --output_video} --output_path {save_video_path}

Author

- LinkedIn

- Blog

- Email: qbxlvnf11@google.com, qbxlvnf11@naver.com