VGG models from ILSVRC 2014
This repository comes with VGG implementation in TensorFlow. VGG took the 2nd place of the ILSVRC-2014 Competition.
Currently, the VGG models in this repo have been tested with CIFAR-10 and CIFAR-100 dataset. As an indivisual deep learner, it is hard to manage such a huge dataset, ImageNet. However, I will keep working on the ImageNet dataset, please wait for it.
VGG16 model example figure from Ref.
VGG: Visual Geometry Group @Oxford University
python vgg.py --model-type ['A'|'A-LRN'|'B'|'C'|'D'|'E'] --dataset ['cifar10'|'cifar100']
import cifar10_utils
import cifar100_utils
from vgg import VGG
...
valid_set = (valid_features, valid_labels)
...
# model type, D is the most well known VGG16 without 1D conv layer
# check the bottom section to see what model types are supported
vggNet = VGG(dataset='cifar10', model_type='D', learning_rate=0.0001)
vggNet.train(epochs=10,
batch_size=128,
valid_set=valid_set,
save_model_path='./model')