interested in contributing in GSOC 2026

[andersendsa]

Hi I am Andersen Dsa , I have been contributing to intel for a while .

I am very interested in contributing and preparing a proposal around it . I wanted to ask if what are the prerequisite tasks for to the following mentors:

-@leoll2 , @kprokofi ( Object Detection Tracking Support for OTX)
would really appreciate to know areas you suggest i could work on before drafting a proposal
Thank you for your time , and i really appreciate any guidance on how to get started on the right foot

[biapalmeiro]

Thanks for your interest @andersendsa. The Open Edge Platform was not formally accepted as an organization for the GSoC 2026 as of now. I suggest wait a bit until having an update from us before investing time in a proposal. Thank you!

[andersendsa]

Thanks for your interest @andersendsa. The Open Edge Platform was not formally accepted as an organization for the GSoC 2026 as of now. I suggest wait a bit until having an update from us before investing time in a proposal. Thank you!

Thank you for the clarification I would love to make any contributions till the intimation comes . Could you please suggest a repo or some good first issues i could contribute too it would be really helpful thanks

[biapalmeiro]

@bharagha @scottmbaker @arodage @saratpoluri @ashwinvijayakumar, @andersendsa is looking for good first issues to start contributing to the Open Edge Platform \o/. Could you please help him to identify those?

You can also check OpenVino GSoC list of projects.

[biapalmeiro]

Hi @andersendsa all contributions to Google Summer of Code will be under OpenVINO this year. Please, check the updates from there: https://github.com/openvinotoolkit/openvino/wiki/Google-Summer-Of-Code

I reached out directly to @leoll2 , @kprokofi, so you might have an answer from them soon.

Good luck! :)

[leoll2]

Hi Andersen, it's great to hear that you're interested in our projects :)
The 'Object Detection Tracking Support for OTX' project description contains a list of prerequisites and preferred skills:

• Python: Intermediate to advanced Python programming skills. Comfortable working with scientific Python libraries (NumPy, pandas) and reading/understanding existing codebases.
• Computer Vision fundamentals: Understanding of object detection concepts (bounding boxes, IoU, confidence scores). Familiarity with tracking basics is helpful.
• PyTorch basics: Experience with PyTorch tensors, basic operations, and working with model outputs.
• Software engineering practices: Ability to write clean, documented code and basic unit tests. Willingness to learn from code reviews.
• Bonus: Prior exposure to tracking algorithms (ByteTrack, DeepSORT, SORT) or video processing.
• Bonus: Experience with deep learning frameworks or production ML systems.

If you would like to delve deeper into one or more of these topics, please let me know what your current level of experience is on that topic (no experience is fine either) and I'll share some resources to learn more about it.

Best of luck!

[leoll2]

Nice, familiarity with PyTorch and multiple contributions to open-source AI-focused projects are definitely a solid start!
Prior experience with object tracking is not a must, no worries about that. If you're curious to explore the topic, here are a few links to get started:

• This and this blog posts by Roboflow. They are well-written and not overly technical.
• These Python examples by Ultralytics, that are relatively easy to setup and play with to familiarize with the concept of multi-object tracking.
• This review paper, which describes the problem, main challenges and mentions several modern approaches. Being an academic paper, it is an advanced reading but it contains many pointers and interesting ideas.

[andersendsa]

Thank you for sharing these resources. I will carefully review them and follow up with you as soon as I can.

[andersendsa]

Hi @leoll2,
I went through all the resources you shared. Thank you for putting them together. I’m planning to start drafting my proposal now, but before I begin, I wanted to check in with you.

Are there any specific guidelines or expectations I should follow before starting the draft? I also wanted to understand what the main focus of this project is. For example, should I prioritize benchmarking rigor, clean integration within OTX, or an optimization-focused approach?

It would also help to know if there is any prioritization among the project ideas so I can align my proposal accordingly.

[leoll2]

Are there any specific guidelines or expectations I should follow before starting the draft?

No specific guidelines other than the generic GSoC ones

I also wanted to understand what the main focus of this project is. For example, should I prioritize benchmarking rigor, clean integration within OTX, or an optimization-focused approach?

I'd say that the most important thing is that the proposal integrates well with the OTX design and API, so that your project can be eventually merged and become part of OTX itself: this would be the ideal scenario for us but also for you as an open-source contributor. To add some context, OTX is the library that powers Geti, one of our flagship projects that enables users to build powerful vision models and deploy them in a streamlined way; although it is not mentioned explicitly in the GSoC project's description to keep the scope limited, one of the long-term goals is to enable object tracking in Geti too: that will be of course easier if OTX supports it natively with a clean interface.
As for benchmarking, that's also important in the final phase to demonstrate the added value of the solution and evaluate the effectiveness of subsequent improvements..

[andersendsa]

Hi @leoll2,

I’ve spent some time mapping out how the project goals like the decoupled design and memory mechanisms would look within the OTX framework, and I wanted to share this diagram as a starting point for my proposal.

• Tracking Post-processor: I’m proposing this as the primary bridge between the OTX Detection Engine and the output. This allows us to keep the detector logic independent.

• State Management: I’ve included a Memory & Appearance Module to store Re-ID embeddings, which I believe addresses the goal of identity stability during occlusions.

• Metric Alignment: The OTX Evaluation Module is shown here as an extension of the existing framework to ensure we can report MOT-specific metrics like IDF1 and MOTA alongside standard detection scores.

Does this modular breakdown align with what you envision for the OTX repo ?