Real-Time Object Counter (RTSP + YOLO + Tracking + Dashboard)

I am looking for a developer with proven experience in Computer Vision to build a real-time counting system using an RTSP stream from an industrial camera. The goal is to replace CountThings and develop our own solution based on YOLO + Object Tracking, capable of counting poultry moving on a conveyor line during the workday.

The system must detect, track, and count each object that crosses a virtual line, display the count in real time, and expose the data in JSON through a REST API / WebSocket for integration into our internal dashboard.

Technical Requirements (Mandatory):
- Real-time video processing via RTSP
- Object detection using YOLO (v8, v9, v10 or similar)
- Object tracking using DeepSORT or ByteTrack
- Line-crossing + ROI logic to prevent false counts
- REST API (JSON) + WebSocket for live counts
- Batch registration (Start/Stop) without video storage
- Simple dashboard (Streamlit / Superset / Metabase / similar)
- Code must be in Python (preferred)
- Stable operation for 8–10 hours per day

Deliverables:
- Full source code (GitHub or similar)
- Docker-ready deployment (preferred)
- Working API (REST + optional WebSocket)
- Real-time dashboard
- Deployment & installation documentation
- Tests using our RTSP stream

Your Proposal MUST Include:
1. Links to similar projects (counting, tracking or RTSP real-time CV systems)
2. Which YOLO model you would use and why
3. Estimated delivery time
4. Budget proposal
5. Suggested improvements (optional)

Estimated Timeline: MVP: 5–10 days | Production-ready version: 2–3 weeks

Required Skills: Python, OpenCV, YOLO, DeepSORT/ByteTrack, RTSP/FFmpeg/GStreamer, experience delivering production-ready CV systems.