zhangming8/ppyolov2-pytorch

Pytorch reimplement of PPYOLOv2, PPYOLO, YOLOv3

Pytorch reimplement of PPYOLOv2

environment

pytorch>=1.7.0, python>=3.6

TODO

add p6, p7 head
release pretrain weight

DONE

EMA
DropBlock
IoULoss
IouAware
GridSensitive
SPP
grad_clip
PAN neck
FPN neck
auto AMP
multi-size training
MixUp
GridMask
mosaic
ResNext, res2net, CSPdarknet backbone
ResNet-vd backbone and add DCN in stage3
warmup
accumulate grad to increace batch size

train

sh train.sh

evaluate

sh evaluate.sh

predict/inference/demo

sh predict.sh

reference

https://github.com/PaddlePaddle/PaddleDetection
https://github.com/open-mmlab/mmdetection
https://github.com/JDAI-CV/fast-reid
https://github.com/ultralytics/YOLOv5
https://github.com/xingyizhou/CenterNet