TorchCAM

TorchCAM

A Docker app from frgfm for class activation maps on PyTorch models, such as CAM and Grad-CAM, which show where an image model looked. For developers who want to see its output in a browser.

Made by frgfm on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: TorchCAM is a Docker-based app from frgfm for class activation maps on PyTorch models. The README line names CAM and Grad-CAM as examples. Such maps show which parts of a picture a model relied on. The Space runs on port 8501 on basic CPU hardware at Hugging Face.

Who it suits: It suits machine learning developers and students who want to see why an image model made a choice, without writing code first.

Industry | Who uses it | How they use it | Example
Machine learning | Model developers | Check what a vision model focuses on | Run a picture and read the activation map
Research | Interpretability researchers | Try activation map methods side by side | Compare the output of two methods
Education | Students | See how explanation maps look | Try a photo of a single object

You can skip it if you want to run the maps inside your own code, since this is a browser demo.

Where to use it:

• A web browser, on the Space's page at Hugging Face
• Your own computer, on Windows, macOS or Linux with Docker

Use case | Who | What they get back
Explain a prediction | Developers | A class activation map for an image
Teach the idea | Instructors | A visual example of model attention
Try the demo locally | Developers | The same app running in a container

How to set it up:

• Nothing to install if you use the Space online
• Python, to download the files with the Hugging Face tool
• Docker, to build and run a copy

1. Open the Space's page on Hugging Face to use it online.
2. To run a copy, install huggingface_hub and download the Space into a folder.
3. Move into that folder and build the Docker image.
4. Run the image with port 8501 published.

pip install -U huggingface_hub
hf download frgfm/torch-cam --repo-type space --local-dir torch-cam
cd torch-cam
docker build -t torch-cam .
docker run -it -p 8501:8501 torch-cam

How to use it: Open the Space and give it a picture. The README gives no steps, so follow the page's own controls. After the docker run command above, open the app in a browser on port 8501 of your own machine.

docker run -it -p 8501:8501 torch-cam

Example: Open the Space online and give it a clear photo of a single object. You should get a map showing which parts of the picture mattered to the model, though the README does not describe the controls.

For newcomers and experts: If you are new, think of the map as a heat overlay that shows where the model looked. Try a photo with one clear subject first.

If you know the field, the Space is a Dockerfile plus a src folder serving an app on port 8501. The README line mentions CAM and Grad-CAM, which are standard class activation map methods. The record also lists several language and vision models whose role in this demo the README does not explain.

You give the app a picture
A PyTorch model classifies it
The CAM method traces where it looked
The page shows the activation map

Good to know:

• License: apache-2
• Cost to run: none named. The online Space runs on Hugging Face, and a copy needs Docker.
• Data it reaches: the pictures you give it, and the models its code pulls from Hugging Face.
• Limits: the README is a single line and gives no list of supported models or methods.

On Hugging Face, 13 November 2026

• Likes: 16
• Built with: docker
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/frgfm/torch-cam
• Files: https://huggingface.co/spaces/frgfm/torch-cam/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.