MaxPolak97/polycam-regression ? reverse-engineered prompt

Build me a Python project for generating and evaluating high resolution saliency maps for image regression models, centered on the PolyCAM methods, PCAM plus, PCAM minus, and PCAM plus minus.

I want to be able to load a pretrained PyTorch image model, pass in a folder of images and a text file with the image list, then generate saliency maps and save them out so they can be viewed later. It should also compute faithfulness metrics like insertion and deletion, save the detailed results to CSV, and make it easy to compare methods side by side. If possible, include a few simple notebooks or a lightweight visual workflow for inspecting the saliency maps and the metric curves.

Please make the scripts easy to run from the command line, with sensible defaults for an ImageNet style setup, and keep the code organized so I can swap in other explanation methods too if the needed libraries are installed. If you need current docs for any helper libraries, look them up online first.