The AP is slower than the paper report,I don't know what caused this.

VisionLearningGroup / DA_Detection

Implementation of "Strong-Weak Distribution Alignment for Adaptive Object Detection"

http://cs-people.bu.edu/keisaito/research/CVPR2019.html

MIT License

349 stars 79 forks source link

The AP is slower than the paper report,I don't know what caused this. #38

Open hustgetlijun opened 3 years ago

hustgetlijun commented 3 years ago

I have run the order : CUDA_VISIBLE_DEVICES=$0 python trainval_net_global_local.py --dataset pascal_vov_water --dataset_t water --net res101 --cudam --lc --gc --epoch 7. the AP is stable as below: bicycle: 0.6945 bird: 0.5345 car: 0.4495 cat: 0.3870 dog: 0.2764 person: 0.624 mean AP:0.4943