Train RetinaNet with Focal Loss in PyTorch.
Reference:
[1] Focal Loss for Dense Object Detection
Imporvements:
Add the eval.py to calculate mAP on VOC.
Support batch_size>1 during test.
Add the produce_txt.py to produce train.txt and val.txt for voc data.
Results: mAP: 69.0 points on VOC2007_trainval datasets.