README.md

Tiny ImageNet 200 Playground

Train and evaluate models with ignite

More detail on the dataset can be found here. Dataset download link

Requirements

We need PyTorch (can be installed from http://pytorch.org/) and the following dependencies:

pip install --upgrade -r requirements.txt

Training

Checkout training configuration, for example, configs/train_vgg16_bn_basic.py and update paths. Next, start training with a simple command:

python tiny_imagenet200_train_playground.py configs/train_vgg16_bn_basic.py

The output folder will contain folders of with training runs:

• training_YYYYmmDD_HHMM
  • train.log : training log
  • 5 best models, model_*.pth
  • 1 last model, checkpoint_*.pth
  • tensorboard logs

TensorboardX training monitoring

tensorboard --logdir=tiny_imagenet200_output

Evaluation

Same as in the training part, edit a conifguration file, for example, configs/test_vgg16_bn_basic.py and run the following script:

python tiny_imagenet200_test_playground.py configs/test_vgg16_bn_basic.py

Predictions are stored in a CSV file in the output folder.