Skip to content

lim0606/caffe-googlenet-bn

Repository files navigation

caffe-googlenet-bn

This model is a re-implementation of Batch Normalization publication, and the model is trained with a customized caffe; however, the modifications are minor. Thus, you can run this with the currently available official caffe version, including cudnn v4 support and multigpu support.

The network definition and solver prototxt files are modified from https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet

Notes:

The uploaded caffemodel is the snapshot of 1,200,000 iteration (30 epochs) using solver_stepsize_6400.prototxt

The uploaded model achieves a top-1 accuracy 72.05% (27.95% error) and a top-5 accuracy 90.87% (9.13% error) on the validation set, using a single center crop.

Thank John Lee for helping me training this model.

Tips for performance

  1. Real-time data shuffling is important
  2. Data augmentation during training should improve the accuracy.
  3. Change interpolation method (default is bilinear) of opencv to bicubic when you convert image will give you minor improvement.

To-do

  1. Data augmentation

References

[1] http://arxiv.org/abs/1409.4842

[2] http://arxiv.org/abs/1502.03167

License

This model is released for unrestricted use.

About

re-implementation of googlenet batch normalization

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages