Skip to content
PyTorch implementation of Local Binary Convolutional Neural Network http://xujuefei.com/lbcnn.html
Python
Branch: master
Clone or download
Latest commit afd3510 Apr 20, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore split train test Sep 3, 2017
README.md mnist toy Apr 20, 2018
lbcnn_model.py mnist toy Apr 20, 2018
main.py mnist toy Apr 20, 2018
requirements.txt mnist toy Apr 20, 2018
utils.py mnist toy Apr 20, 2018

README.md

Pytorch implementation of CVPR'17 - Local Binary Convolutional Neural Networks.

Training even MNIST with parameters, stated in the original repository, is incredibly slow. Here is an example of training a toy model -- "2 x {BatchNorm2d(8) -> ConvLBP(8, 16, 3) -> Conv(16, 8, 1)} -> FC(200) -> FC(50) -> FC(10)" -- on MNIST:

Epoch 0/5: 100%|██████████| 235/235 [00:06<00:00, 37.74it/s]
Epoch 0 train accuracy: 0.948
Epoch 1/5: 100%|██████████| 235/235 [00:05<00:00, 41.98it/s]
Epoch 1 train accuracy: 0.962
Epoch 2/5: 100%|██████████| 235/235 [00:05<00:00, 42.01it/s]
Epoch 2 train accuracy: 0.969
Epoch 3/5: 100%|██████████| 235/235 [00:05<00:00, 42.04it/s]
Epoch 3 train accuracy: 0.971
Epoch 4/5: 100%|██████████| 235/235 [00:05<00:00, 41.84it/s]
Epoch 4 train accuracy: 0.971
Finished Training. Total training time: 41 sec
Full forward pass: 100%|██████████| 40/40 [00:00<00:00, 100.42it/s]
MNIST test accuracy: 0.974
You can’t perform that action at this time.