Skip to content
Pytorch re-implementation of Fssd
Branch: master
Clone or download
Latest commit 7eb06bd Mar 19, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data fssd on pytorch0.4 Dec 9, 2018
image fssd on pytorch0.4 Dec 9, 2018
layers fssd on pytorch0.4 Dec 9, 2018
models BN for FSSD Mar 18, 2019
utils fssd on pytorch0.4 Dec 9, 2018
README.md Update README.md Mar 19, 2019
demo.py fssd on pytorch0.4 Dec 9, 2018
make.sh fssd on pytorch0.4 Dec 9, 2018
slim_prune.ipynb add prune file Mar 18, 2019
test.py modify for prune Mar 18, 2019
train.py Update train.py Mar 19, 2019

README.md

Fssd.pytorch

A Pytorch0.4 re-implementation of Fssd detection network, with pretrained weights on VOC0712 and mAP=79.74.

Network train data test data mAP FPS Download Link
FSSD VOC 07+12 VOC 07 test 79.74 90 Baidu/Google

Official caffe implementation is here, and pytorch0.3 re-implementation is here.

For details, please read the paper: FSSD:Feature Fusion Single Shot Multibox Detector

Our re-implementation is slightly better than paper (79.74 vs 78.8). Inference time is tested on TITAN X Pasal and cudnn V5. More high performance models will be available soon.

About more details

  • The implementation code abandons the bn layer after feature fusion.
  • Because the limitation of amount of gpu, batch_size is set to 32. If use 64 or more, I believe it will produce better performance.
  • If you want to re-produce more accurate model, I believe that adding BN to the body of FSSD and applying Mixup will be very effective.
  • The code is mainly based on rfbNet. If you are interested in this project, please email me(yhao.chen0617@gmail.com)

Update

I add a prune file for FSSD, which bases on Network Slimming (ICCV 2017), for details, please refer to the paper: Network Slimming.

State mAP FPS GPU Model Size Download Link
Original 79.74 90 TITAN X Pascal 136M Baidu/Google
First Prune (50%) 79.64 150 TITAN X Pascal 52M Bd(p6fe)/Gg
Second Prune

Steps for pruning the model:

  • Training the FSSD with BN model, meanwhile applying L1 constraints to the parameters of BN
  • Applying Network Slimming to the model trained in the previous step (Prune threshold is defaulted to 50%)
  • Re-train the pruned model, approximate 200 epochs, drop 0.1% in accuracy.
  • If still prune the model, I believe it can produce the result of mAP>79%, FPS~200. Of course, for the pruned model, BN can be merged to the convolution layer to be further accelerate.
You can’t perform that action at this time.