convet-study: A study of ConvNet architectures on small image datasets

This repository contains implementations of some architectures that I find interesting and get good results on small image datasets (mostly CIFAR-10).

Requirements

This code runs on Keras and has been tested with Tensorflow as backend. You should have a GPU to run this code on a reasonable amount of time.

• Python 2.7.
• Numpy.
• Scipy.
• Pyyaml.
• Tensorflow.
• Keras.
• hdf5, h5py.

Getting the data

Download the code, run the get_dataset.sh script, it should download the datasets and place them on the appropriate folders.

Training a model using the train.py script

This script allows training (some) models using a command-line interface. You should pass at least a model loader and a dataset, all the other arguments are optional and have default values. Assuming you are in an ipython session, your call should be something like

%run train.py -d DATASET -m MODEL_LOADER -b BATCH_SIZE --l2 L2_REGULARIZATION -s SAVE_FILE_NAME --schedule LEARNING_RATE_SCHEDULE --lr LEARNING_RATE -e NUMBER_OF_EPOCHS -v VALID_RATIO

This script will automatically save the model at every epoch using keras model saving (SAVE_FILE_NAME.h5) along with a metadata file (SAVE_FILE_NAME.meta) that contains the arguments passed, training metrics and learning rate scheduling information. Should you want to resume training run the script with the -l[--load] argument

%run train.py -l SAVE_FILE_NAME

This will resume training from the last checkpoint.

Training a model using your own script

All the models are defined by a function in the models package. Calling it will return a Keras model that can be trained however you want. You can check the scripts directory for examples.

Implemented models

Densenet

Implementation of [1]. Currently only tested the smaller densenet: 40 layers total, 3 dense blocks with 12 layers with and growth rate of 12. The code should support the bigger models though. Accuracy of 93.58% without data augmentation (paper reports 93%) and 94.72% (paper reports 94.76%) with horizontal flips and crops.

Batch-normalized Network in Network

Architecture based on [2], modified to include batch normalization. Achieved 90.97% accuracy without data augmentation and 92.46% with flips. Uses learning rate schedule from [3].

Batch-normalized VGG-like network

Bases on [3], 91.68% accuracy without data augmentation and 92.92% with flips.

References

• [1]: G. Huang, Z. Liu and K. Weinberger. Densely Connected Convolutional Networks. Arxiv. Github.
• [2]: M. Lin, Q. Chen and S. Yan. Network in Network. Arxiv.
• [3]: S. Zagoruyko. cifar.torch. Github.