Using keras2.2.4 to pruning VGG16
This is a project using keras to prune CNN, now I finish it and wanna share.
Final result: acc：0.931 -> 0.876 model size：57.3 M -> 5.85M.
The weight file is logs/weights/ iter_3_prune_trained_model_weights.h5, and the model is logs/weights/iter_3_prune_trained_model_json.json.
The final model is in the model.py, you can change it by yourself.
Please read readme.txt first, thera are what i want to speak.
Reference：
https://github.com/jacobgil/pytorch-pruning
https://github.com/geifmany/cifar-vgg
https://github.com/BenWhetton/keras-surgeon
Any questions please connect to me : QQ 997398715