Color Quantization of CIFAR-10 Image Dataset for Image Classification using CNNs

TJHSST Computer Systems Research Lab with Dr. Selma Yilmaz

Applied vector quantization and Floyd-Steinberg dithering to reduce the computational load of image recognition. Maintains high classification accuracy while improving efficiency. Aiming to optimize neural network inputs for low-power devices like surveillance and thermal cameras. Final result is a complete color-quantized version of CIFAR-10 with public licensing for use.

How to Run Code

After cloning and cd into the repository, run the CIFAR10-CNN.py and wait a few minutes for training to begin.
Quit the program at any time to view training progress across epochs. To resume, run the python file again and follow instructions in the terminal.