neelu5/pytorch-transfer-learning-of-vgg19-for-cifar-10-dataset

PyTorch Transfer Learning of VGG19 on Cifar-10 Dataset

Transfer Learning of VGG19 trained using ImageNet an retrained for Cifar-10 Dataset using PyTorch.

Requirement

Library	Version
Python	^2.7
PyTorch	^0.1.12
Numpy	^1.12.0
Pickle	*

Usage

Download code:

git clone https://github.com/eugenelet/PyTorch-Transfer-Learning-of-VGG19-for-Cifar-10-Dataset

cd PyTorch-Transfer-Learning-of-VGG19-for-Cifar-10-Dataset

Train cnn:

Batch size: 128

Prediction made on per epoch basis.

161 epochs takes about 3h on GTX 1080.

Retrain Model:

python train.py

Make prediction:

Using Model Zoo

python test.py

Using vgg19.npy

python test_vgg19.py

Tensorboard

tensorboard --logdir=./tensorboard

License

Apache License 2.0

Implementation Details

My Blog