This repo is the official implementation of CALD: Consistency-basd Active Learning for Object Detection, accepted to the Workshop on Learning With Limited Labelled Data for Image and Video Understanding (L3D-IVU), in conjunction with CVPR 2022.

• pytorch>=1.7.1
• torchvision=0.8.2

(option if you want to get class-wise results of coco)

• mmcv=1.0.4
• pycocotools=2.0.2
• terminaltables=3.1.0

python cald_train.py --dataset voc2012 --data-path your_data_path --model faster

If you are interested with our work, please cite this project.

@article{yu2021consistency,
  title={Consistency-based active learning for object detection},
  author={Yu, Weiping and Zhu, Sijie and Yang, Taojiannan and Chen, Chen and Liu, Mengyuan},
  journal={arXiv preprint arXiv:2103.10374},
  year={2021}
}