This project is done based on the result from the final research project in the CSE163 course from the University of Washington about Using different models in Image Classification. The link below gives more information about that research project: https://github.com/quocthai9120/Images-Classification.
The previous result ended up with 70.09% correct accuracy for 10000 testing instances using the CIFAR_10 dataset. This is not a bad start for the image classification task. However, we need to improve the model significantly if we want to use in real-life classification tasks.
This project focuses on experimenting with different methods to improve the Convolutional Neural Network model that used to image classification using the CIFAR_10 dataset, including data preprocessing, pre-trained model modification, and training modification.