In this homework, you are tasked with building a convolutional neural network (CNN) for image classification, potentially incorporating advanced training techniques. You are allowed to use a pretrained model and fine-tune it on the provided dataset. However, using external datasets or pretrained models for training purposes is prohibited.

Start by building a simple CNN as a baseline. Then, you can explore improving the model by designing a more effective architecture or experimenting with different data augmentation strategies to enhance performance.

The dataset consists of 20 food categories for classification, with 7,500 training samples, 2,000 validation samples, and 4,000 testing samples.