⚠️ ⚠️ [ READ APPLICATION PROTOCOL] ⚠️ ⚠️
Kornia-rs GSoC 2026 Project Ideas
Timeline (official):
- Org application deadline: Feb 3, 2026 (18:00 UTC)
- Accepted orgs announced: Feb 19, 2026 (18:00 UTC)
- Contributor application window: Mar 16–Mar 31, 2026 (18:00 UTC)
How this ideas list helps our application:
- Clear, scoped projects with measurable outcomes
- Defined prerequisites and evaluation criteria
- Community impact across core kornia-rs crates
AI Tooling Policy (GSoC 2026)
We follow Google’s 2026 guidance on AI tooling for both contributors and mentoring orgs. Using AI is allowed only when it is transparent, limited, and fully understood by the contributor. Raw AI output is not accepted.
Allowed uses (with disclosure):
- Research and learning (reading papers, summarizing docs)
- Boilerplate/refactoring, test scaffolding, or debugging support
Not allowed:
- Submitting AI‑generated proposals as-is
- Committing code you cannot fully explain
- Copying AI output without review and verification
Contributor responsibility: You remain 100% responsible for correctness, licensing, and explaining every line you submit. If you used AI, say where and how in your proposal and PRs.
Projects Index
- GPU acceleration for kornia-tensor and kornia-imgproc
- Expand kornia-vlm with ONNX/TensorRT backends
- kornia-SLAM: baseline Visual-Inertial Odometry (VIO)
- Bubbaloop simulation integration and improvements (MuJoCo)
- Bubbaloop real-robot integration
Projects are ordered by soft priority.
1. GPU acceleration for kornia-tensor and kornia-imgproc
Description Implement GPU kernels for core tensor ops and image processing transforms in Rust. Focus on frequently used operations (resize, warp_affine/warp_perspective, color transforms, distance transform) and ensure API parity with existing CPU implementations. Use CubeCL (multi-platform GPU compute in Rust) and/or native CUDA interop as appropriate.
Expected Outcomes
- Minimal GPU backend for kornia-tensor ops used by kornia-imgproc
- GPU implementations for a handful of high-impact transforms
- Benchmarks vs CPU and basic docs/examples
Resources
- CubeCL (Rust GPU compute): https://github.com/tracel-ai/cubecl
- Rust std on GPU (Vectorware blog): https://www.vectorware.com/blog/rust-std-on-gpu/
- CUDA Toolkit docs: https://docs.nvidia.com/cuda/
- kornia-rs repository: https://github.com/kornia/kornia-rs
Possible Mentors: Edgar Riba, Jian Shi, Christie Purackal
Difficulty: Hard
Duration: 350 hours
2. Expand kornia-vlm with ONNX/TensorRT backends
Description Bring more vision-language models from kornia (PyTorch) into kornia-rs and deliver optimized inference via ONNX Runtime with TensorRT where available. Focus on model portability and clear runtime selection (CPU/CUDA/TensorRT), with an emphasis on embedded targets (e.g., Jetson).
Expected Outcomes
- 1–2 VLM models integrated into kornia-vlm
- ONNX Runtime backend + optional TensorRT path
- Benchmarks and a small demo
Resources
- ONNX Runtime TensorRT Execution Provider: https://onnxruntime.ai/docs/execution-providers/TensorRT-ExecutionProvider.html
- ONNX-TensorRT backend: https://github.com/onnx/onnx-tensorrt
- Kornia (PyTorch) project: https://github.com/kornia/kornia
- kornia-rs repository: https://github.com/kornia/kornia-rs
Possible Mentors: Edgar Riba, Miquel Farré
Difficulty: Medium–Hard
Duration: 350 hours
3. kornia-SLAM: baseline Visual-Inertial Odometry (VIO)
Description Formalize and extend our kornia-SLAM crate within kornia-rs and implement a baseline visual-inertial odometry pipeline. Extend the crate to support stereo VIO, add GPU support, optimize existing kernels, and extend functionality toward LiDAR SLAM techniques. Emphasis on a modular API that can later host learning-based modules.
Expected Outcomes
- kornia-slam crate integrated into workspace
- Baseline VIO pipeline + dataset evaluation
- Benchmarks for CPU/GPU performance and accuracy
- Documentation + example CLI
Resources
- SLAM Handbook: https://github.com/SLAM-Handbook-contributors/slam-handbook-public-release
- ORB SLAM3 repository: https://github.com/UZ-SLAMLab/ORB_SLAM3
Possible Mentors: Christie Purackal, Dmytro Mishkin, Edgar Riba
Difficulty: Hard
Duration: 350 hours
4. Bubbaloop simulation integration and improvements (MuJoCo)
Description Improve the Bubbaloop simulation stack by integrating or enhancing a MuJoCo-based workflow. Focus on stable simulation bindings, reproducible training/evaluation loops, and better tooling for dataset generation and benchmarking.
Expected Outcomes
- MuJoCo integration with example tasks
- Simulation-focused training/evaluation scripts
- Documentation + demo video
Resources
- Bubbaloop repository: https://github.com/kornia/bubbaloop
- kornia-rs repository: https://github.com/kornia/kornia-rs
Possible Mentors: Edgar Riba, Jian Shi, Christie Purackal
Difficulty: Medium–Hard
Duration: 350 hours
5. Bubbaloop real-robot integration
Description Connect the Bubbaloop learning framework to real robot hardware. Focus on hardware abstraction, safety/recovery procedures, and a minimal real-world demo that mirrors a simulation task.
Expected Outcomes
- Hardware integration layer with a small example robot
- Real-world demo aligned with a simulation scenario
- Documentation + setup guide
Resources
- Bubbaloop repository: https://github.com/kornia/bubbaloop
- kornia-rs repository: https://github.com/kornia/kornia-rs
Possible Mentors: Edgar Riba, Jian Shi, Christie Purackal, Miquel Farré
Difficulty: Medium–Hard
Duration: 350 hours
General Evaluation Criteria
- Code quality: idiomatic Rust, clear APIs, tests, docs
- Performance: benchmarks and regression tracking where relevant
- Reproducibility: scripts, datasets, and/or instructions to validate results
- Impact: usefulness to kornia-rs users and downstream projects
Prerequisites (Required for Consideration)
To reduce spam proposals and prioritize contributors who can execute, applicants must complete all of the following before submitting a final proposal:
- Join our community chat and introduce yourself
- Make at least one small, accepted PR to kornia-rs (or related crate)
- Share a draft proposal with a mentor for feedback
- Be able to explain the code you submitted (short interview or async review)
Selection Process (How we evaluate)
We select contributors, not proposals. Proposals help shortlist, but we prioritize:
- Engagement in the community and responsiveness to feedback
- Demonstrated ability to explain code and design choices
- Evidence of learning and independent problem‑solving
Pre-Application Work (Recommended)
- Read the application protocol and join the community channels
- Pick a project and submit a small PR to kornia-rs
- Write a short design proposal with milestones and risks