Created the concept of normalizer
- Average Pooling - VGG 16 mod
- Implemented the batch normalization - Removed the hdf5 backend
Introduced the VGG16
New learning rate
A trainer can load a pretrained network
New FaceNET Pooling with strides
Fixed FaceNET
New load and save methods
Update learning rate issue #7
Fixed problem with the learning rate
Managing better the GPUs
New triplet selection
Triplet selection Data augmentation
Organized the trainer
Tests with triplet
Chopra network triplet nets
Testing Siamese net