Computer Vision / Backend Engineering
FleetRobo
FleetRobo — Driver Face Recognition System
A driver recognition system developed during an industry internship, combining face identification, unknown-person rejection, image/video processing, and embedding-based similarity search.
Project overview
An industry-internship project focused on backend services and computer-vision workflows for truck-driver recognition.
- Python
- FastAPI
- PostgreSQL
- OpenCV
- FAISS
- InsightFace
Problem & internship context
FleetRobo was developed during an internship at Binary Semantics Ltd. for a driver-recognition context where identification quality and handling unfamiliar faces both matter.
Recognition architecture
The system combines face detection and recognition with embedding comparison, image and video processing workflows, and quality checks.
Known / unknown classification
Unknown-person rejection is a core requirement, separating recognised drivers from inputs that should not receive a known identity.
Similarity search & APIs
FAISS supports embedding-based similarity search. FastAPI endpoints and PostgreSQL integration provide the backend layer around recognition workflows.
Recognition quality
Recognition quality checks and unknown-person rejection are important parts of the workflow. Image and video processing introduce different input conditions, making it important to distinguish a useful match from a face that should remain unrecognized.
Next project