This project involved building and training a convolutional neural network (CNN) from scratch using PyTorch and the CIFAR-10 image dataset. The CNN model is defined by 4 convolution layers with batch normalization, ReLU, dropout, and max pooling operations, followed by a linear layer and a classifier layer at the end.
The model was trained with the 50,000 images training set for the CIFAR-10 dataset and evaluated with the 10,000 images testing set. During training, the model was subjected to training for 25 epochs and was able to attain a 65% classification accuracy metric on the training images. For testing, the model was able to attain a 62% classification accuracy metric on the testing images.
Software: Python | PyTorch