GSoC 2026 - Interest in Object Detection Tracking Support for OTX

[om4rrr]

Hi @kprokofi @leoll2

I'm Omar Ali, a CS Student. I'm very interested in the Object Detection Tracking Support for OTX project for GSoC 2026.

I'm fixing an issue where negative values were not being checked in properties with unsigned value types: PR #33482

I'm currently starting to work on my proposal and need some guidance:

Are there specific parts of the OTX codebase I should focus on first?

Any advice or resources to help me write a strong proposal would be greatly appreciated!

[leoll2]

Hi @om4rrr, happy to see you interested in the project :D
If you haven't already, I suggest taking a look at this discussion, it may answer some of your questions.

Are there specific parts of the OTX codebase I should focus on first?

Since the codebase is currently being restructured and there are some moving parts, I'd suggest to use the release-2.6 branch as a reference; the public docs are also more or less aligned with that branch.

[om4rrr]

Hi @leoll2
I have submitted my proposal on the GSoC dashboard and I'm excited to get your feedback.

For the next period I want to be more active and I'd like to work on some tasks/issues related to the project.

Could you assign tasks/issues to me?

[leoll2]

Thanks for your submission! At the moment we don't have 'good first issues' in our repo, however you could work on one of the following task as a trial:

• create a Jupyter notebook that trains an OTX model of your choice, applies a detector-agnostic tracker on the output (e.g. DeepSORT) and finally shows the predictions, with and without the detector, on a demo input video.
• integrate a new model architecture of your choice in OTX, train it on a small/medium dataset and benchmark it against another existing model