CIFAR-10 Benchmark CNN

Re-implementation of 94% on CIFAR-10 in 3.29 Seconds on a Single GPU for TAMU's CSCE-636: Deep Learning project.

Report: report.pdf

Steps to run code

Given the data is in 'data' directly, outside of 'code' directory.

Training

python3 main.py train ../data .

Testing

cd code
python3 main.py test ../data .

Predicting

cd code
python3 main.py predict ../data ../predictions

This creates a predictions.npy file outside of 'code' directory.

Directory Structure

|- code (directory containing all the python code files)
|- data (directory containing all the training and testing CIFAR-10 data)
|- saved_models (directory containing the saved model which can be used for testing/predicting)
|- logs (directory containing logs for all experimentation runs) |- README