Computer Vision Engineer – Edge-Based Human Detection & Alerting System

Project Description We’re hiring a Computer Vision Engineer to assist with building a lightweight, edge-deployable vision system that detects human presence events from CCTV camera feeds and triggers notifications based on stable detection logic. The emphasis is on engineering reliability, not experimentation or academic exploration. The system should be designed to operate continuously and handle real-world noise gracefully. Scope of Work • You will help implement a computer vision pipeline that: • Connects to RTSP-based CCTV streams • Detects human presence entering a defined outdoor area • Triggers a single event notification per entry • Uses logic to avoid duplicate or noisy alerts • Operates as a long-running background process Key Constraints (Please Read Carefully) • No proprietary footage will be provided • Testing must be done using: • Publicly available videos • CCTV-style sample footage • Internet-sourced material representative of outdoor scenes • Camera viewpoints are not fixed • Expect one or two general camera perspectives, but nothing scene-specific • The solution must be generic and reusable, not tuned to a single setup Environmental Challenges to Account For • The system should behave sensibly when: • People are only partially visible (upper body, legs, silhouettes) • Individuals move between parked vehicles or obstacles • Temporary occlusion occurs and people reappear • Background motion causes brief detection instability • Multiple people are present at the same time • Lighting, shadows, reflections, and clutter are present • Only specific areas of the frame should be monitored using zone-based masking Accuracy at every frame is less important than consistent event-level behaviour. Event Detection Behaviour The logic should: • Detect a new arrival event when a person enters a monitored zone • Avoid repeatedly triggering events for the same individual • Handle multiple people entering independently • Function without requiring the monitored area to reset or clear Engineering Expectations The implementation should: • Run as a persistent process • Restart automatically if interrupted • Recover from camera stream interruptions • Log key events and errors • Use simple configuration files for streams, zones, and thresholds Deliverables • Functional human presence detection pipeline • Zone / ROI masking support • Stable event-trigger logic • Logging and basic health checks • Setup and usage documentation Who This Is For • This task is best suited for engineers who: • Have worked on real-world computer vision systems • Understand the limitations of detection models in uncontrolled environments • Can design robust logic around imperfect detections • Are comfortable making trade-offs for edge-based deployments Budget • Fixed budget: up to $150 • This is a clearly scoped task, not an open-ended engagement When Applying, Please Include • Your approach to human detection and event triggering • How you would handle occlusion and intermittent detections • How you would prevent repeated or false triggers • How you would implement zone-based monitoring • Examples of relevant past work (production or applied projects preferred) Applications that do not address these points will not be considered. When an entry event is confirmed, the system must send a WhatsApp notification to a predefined dealership phone number using a programmatic WhatsApp integration. Apply tot his job