The document introduces a computer vision XAI project. It discusses implementing several attribution and attention methods on classification models for MNIST and CIFAR10 datasets, including Convolutional Block Attention Module (CBAM), Class Activation Map (CAM), and Residual Attention Network (RAN). Over 400 training runs will be performed to evaluate the different methods on base models and retrained models with salient regions removed or kept. The project files are organized to load data, define models, and visualize saliency maps and evaluation results.