This is the HuggingFace community for the SAFE: Image Edit Detection and Localization Challenge 2025.

The challenge is hosted by the UL Digital Safety Research Institute (DSRI) and is co-located with the SynRDinBAS: Synthetic Realities and Data in Biometric Analysis and Security Workshop @ WACV 2026.

How to participate

To participate in the challenge, you need to do three things:

1. Visit the challenge home page and sign up using the linked registration form. After verifying your team's email, you will receive access credentials for the submission platform.
2. Implement your detector model. You can use the example-submission repository as a starting point, but you don't have to.
3. Submit your detector model for evaluation. You can build your submission package yourself and submit it using a CLI tool (preferred), or you can build your submission in a HuggingFace Space and submit the Space using a web form.

How to make a submission

The infrastructure for the challenge runs on DSRI's Dyff platform. Submissions to the challenge must be in the form of a containerized web service that serves a simple JSON HTTP API.

If you're comfortable building a Docker image yourself, the preferred way to make a submission is to upload and submit a built image using the Dyff client.

Alternatively, you can create a Docker HuggingFace Space and create submissions from the space using a webform. The advantage of using an HF Space is that it builds the Docker image for you. However, HF Spaces also have some limitations that you'll need to account for.

Getting started

Check out the example-submission model and the pilot-1 dataset repos.

The example-submission demonstrates how submissions for the challenge are structured. You will package your detector model as a containerized web service that serves a simple JSON API. You will then upload your detector to DSRI's testing platform, where it will be evaluated on private datasets.

The pilot-1 dataset is an example of the data format you can expect in the challenge. The pilot task data are public and scores on pilot tasks will not count toward challenge rankings.

⚠️ Important notes ⚠️

⚠️ Submissions will run without Internet access

For security reasons, your submitted detector will be blocked from accessing the public Internet. Your submission package must include all necessary files. Make sure your system does not attempt to download model files at run-time, and note that downloading on demand is the default behavior of many popular ML packages like transformers and huggingface_hub. You can test that your system starts successfully without Internet access by using a command like:

docker run --network none ...

⚠️ CUDA version compatibility

Submissions that request a GPU will be allocated 1x Nvidia L4. L4 GPUs require CUDA 11.8 or higher. Our system is running CUDA driver version 535.

You must include compatible CUDA Toolkit libraries in your submission's Docker image.

We recommend using nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04 or a similar image as your base image. This image has been validated to work correctly.

⚠️ CUDA 12.8 and CUDA 13.x are not supported by our infrastructure

You will see an error like the following:

RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx

We recommend using nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04 or a similar image as your base image. This image has been validated to work correctly.