المسؤوليات والمهام الرئيسية:

• Design and maintain real-time CV pipelines (capture → infer → track → analyze → emit

events)

• Train & fine-tune YOLO-based detection models (PPE, people, objects, behavior)

• Implement ROI-based logic, line-crossing, and temporal filtering

• Integrate multi-object tracking (ByteTrack / DeepSORT)

• Optimize inference for edge devices (Jetson Nano/Xavier/Orin)

• Convert & deploy models using ONNX → TensorRT

• Handle camera instability, occlusion, lighting variance, and real-world noise

• Expose CV outputs via REST APIs for dashboards & SaaS platforms

• Improve FPS, latency, memory usage, and model robustness in production

 

Technical Stack:

Models: YOLO v8 / v11, Pose Estimation, InsightFace

ML: PyTorch, Ultralytics

CV: OpenCV, Supervision

Tracking: ByteTrack, DeepSORT

Streaming: RTSP / HLS

Backend: Python, Flask, SQLAlchemy

Edge: NVIDIA Jetson, TensorRT, ONNX

DevOps: Docker