YOLO Computer Vision Engineer for Real Time Worker Detection in a Factory

Job Description: We are looking for an experienced Computer Vision and Machine Learning engineer to build a real-time system that detects and tracks people working on our factory floor. The system will run on live video from our 4 existing IP cameras and must perform reliably in an industrial environment with changing lighting, machinery blocking parts of the view, and workers wearing uniforms and safety gear. About the Project We run a manufacturing facility and want to use AI to understand worker presence and activity across the production floor. Each of our 4 cameras covers a different area, including production lines and workstations. The goal is to automatically detect every worker in view, track them over time, and turn that into useful data and alerts for floor management. Scope of Work Review our camera setup and sample footage, then recommend the best technical approach Build a training dataset from our footage and annotate it properly Fine-tune a YOLO model (YOLOv8, YOLOv11, or newer) for accurate person detection in our specific environment Add multi-object tracking (ByteTrack, BoT-SORT, or similar) so each worker keeps a consistent ID across frames Build zone-based logic that can: Count how many workers are present in each defined area Detect when a workstation is left unattended for a set amount of time Measure how long workers spend in each zone Detect basic PPE such as helmets and safety vests Connect the system to all 4 live RTSP camera streams running at the same time Optimize the pipeline to run smoothly in real time on a single local GPU machine Build a simple web dashboard showing live camera views with detections, worker counts per zone, alerts, and daily summary reports with CSV export Deliver clean, well-documented code and walk us through the full system Deliverables Trained YOLO model weights with evaluation results (precision, recall, mAP) Complete source code with clear documentation Real-time inference pipeline for 4 simultaneous camera streams Easy way to draw and edit zones for each camera Web dashboard with live view, alerts, and reports Step-by-step setup and deployment guide Video call handover session to explain and test everything together Required Skills and Experience Strong hands-on experience with YOLO models (Ultralytics YOLOv8 or later) Proven work on person detection and multi-object tracking in real-world video Strong Python skills with PyTorch and OpenCV Experience working with RTSP and IP camera streams Experience with annotation tools such as Roboflow, CVAT, or Label Studio Knowledge of model optimization for deployment (TensorRT, ONNX) Ability to handle difficult conditions like low light, crowded scenes, and partial occlusion Nice to Have Previous projects in manufacturing, industrial safety, or workplace monitoring Experience with PPE detection models Experience with pose estimation for activity recognition Experience building dashboards with Streamlit, FastAPI, or React Experience deploying on NVIDIA Jetson devices How to Apply In your proposal, please include: Examples of similar computer vision projects you have built, with links or demo videos if possible Which YOLO version and tracker you would choose for this project and why The GPU or hardware you would recommend for running 4 camera streams in real time Your estimated timeline to complete the project We will prioritize freelancers who show real deployed work rather than only notebooks or tutorials. Screening Questions Describe a real-world YOLO project you deployed. What accuracy and FPS did you achieve? How would you handle workers being partially hidden behind machines? How would you keep the same ID for a worker who leaves the camera view and comes back? What is your experience running multiple camera streams on one machine?