Objective

• Build a lightweight Video Management System (VMS) dashboard enabling multi-camera visualization, event tagging, video archiving, and intelligent search based on metadata (e.g. person/vehicle/color).
• Deliver a prototype that demonstrates semantic search capabilities (for example: “find all red vehicles between 14h–16h”) and a storage strategy supporting motion-, time- or object-based recording.

Key Components & Responsibilities

• Implement a multi-camera viewer UI supporting simultaneous streams and event-driven playback controls.
• Design and implement storage strategies (motion-based, time-based, or object-based recording) and video archiving policies.
• Develop metadata extraction pipelines to generate labels, timestamps and object types from video (person, vehicle, color, etc.).
• Implement semantic search over indexed metadata to support complex queries (time ranges, object attributes, combinations).
• Tagging and event workflow: allow tagging of events, linking tags to archived clips, and enabling filtered playback/search results.

Technical Environment & Tools

• Backend: Python with FastAPI for APIs and metadata handling.
• Frontend: JavaScript (React.js) for the multi-camera dashboard and user interactions.
• Computer vision / inference: OpenCV plus either DeepStream SDK or OpenVINO for metadata extraction (object detection, color classification, tracking).
• Metadata indexing / storage: Elasticsearch or PostgreSQL for metadata indexing and query support.
• Deployment: Docker Compose for local/containerized deployment and reproducible stacks.
• Optional integrations: support for Milestone/ONVIF APIs to connect to professional VMS deployments if required.

Expected Deliverables

• A working dashboard demo with multi-camera viewing, event tagging and playback of archived video segments.
• A metadata extraction pipeline producing searchable labels and timestamps, and an index enabling semantic queries.
• Documentation covering architecture, deployment (Docker Compose), how to run the demo, and recommendations for production integration (e.g., ONVIF/Milestone).
• Tests or evaluation notes demonstrating accuracy/limitations of the chosen inference stack (OpenVINO or DeepStream) and storage strategy impacts.

How to Apply

• Apply via the company website: https://www.hydatis.com
• Or send your application by email to stages@hydatis.fr with the subject line: "Application Sec-02 - Video Management Dashboard PFE".
• Include CV, cover letter describing relevant experience (computer vision, React/FastAPI, Docker) and any links to prior projects or code repositories.