Advanced Real-Time Visual Object Tracking System (Linux, High-Performance)

Project Description

We are looking for an experienced Computer Vision engineer to develop a high-performance real-time object tracking system running on Linux.

The system will allow a user to select a target within a live video stream and maintain robust tracking under dynamic conditions.

This is not a basic OpenCV demo project. We require a stable, production-oriented architecture with strong tracking persistence under motion, scale variation, partial occlusion, and illumination changes.



Core Functional Requirements
• Linux-based implementation (Ubuntu preferred)
• Real-time video stream processing (USB / CSI / RTSP compatible)
• User-initiated ROI selection (click-to-track)
• Persistent target tracking at minimum 30 FPS (hardware dependent)
• Continuous output of:
• Bounding box
• Target centroid (X,Y)
• Confidence metric (if applicable)



Performance Expectations

The tracker must:
• Handle rapid target motion
• Adapt to scale and orientation changes
• Maintain lock under partial occlusion
• Recover gracefully if tracking confidence drops
• Avoid drift over time

A re-detection or hybrid tracking strategy is preferred if it improves robustness.



Technical Requirements

Preferred stack:
• Python + OpenCV OR C++ + OpenCV
• Modular architecture
• Hardware acceleration support (CUDA / TensorRT) is a strong plus
• Experience with:
• Siamese-based trackers
• DeepSORT-like approaches
• Hybrid detection + tracking pipelines

Clean, well-documented code is mandatory.



Deliverables
1. Fully functional Linux application
2. Source code repository
3. Setup instructions + dependency list
4. Short demo video
5. Optional: performance benchmark report (latency / FPS)