About

I’m looking for a developer to create a proof-of-concept (PoC) version of a dual-camera video stitching system. The goal is to combine two overlapping MP4 recordings (approximately 90 minutes each, 1080p–4K) into a single panoramic output that is visually stable and consistent. This is an early prototype phase. The objective is to demonstrate feasibility and performance, not to produce a final or fully automated system. If the results meet expectations, there will be opportunities for further paid milestones involving AI-based tracking and highlight extraction. Stitch two fixed-camera videos into one panoramic output. Implement basic per-chunk re-alignment (dynamic correction) to prevent drift over long videos. Apply basic colour and exposure correction to match both feeds. Use GPU acceleration (CUDA/NVENC) for efficient 4K processing. Provide a Dockerized implementation that can run on cloud GPU instances (AWS, Azure, or similar). Deliverables Source code or working script (language and libraries flexible). Dockerfile or container image ready for GPU runtime. README with setup, usage, and sample commands. Example stitched output demonstrating correct alignment and blending. Technical Notes You may use any efficient toolset you’re comfortable with. Python + OpenCV + FFmpeg is acceptable, but alternative stacks (C++, GStreamer, CUDA kernels, etc.) are welcome as long as they meet the performance and quality goals.