Featured: Computer Vision (2 projects)

Quick demos + what I built. Full code in linked repos.

Repo: https://github.com/ary-ban/edge-vision-swarm-demo

• Two “agents” run YOLOv8 (ONNXRuntime) and send detections to a Coordinator.
• Hungarian assignment deconflicts targets; center “nudge” logic for smooth handoff.
• ONNX export + optional INT8 quant; real-time overlays.

Tech: Python, OpenCV, ONNXRuntime, SciPy, UDP sockets
What I built: UDP JSON protocol, cross-agent fusion on shared X-axis, Hungarian assignment, overlays, ONNX/INT8 tools.

Repo: https://github.com/ary-ban/FaceRecog

• Tried multiple backbones (Facenet, ArcFace, dlib); picked InsightFace/ArcFace after papers + trials.
• Detection fixed with YOLOv8; embeddings stored in SQLite; cosine similarity search.
• Includes pricing model spreadsheet and notes on when vector DBs make sense.

Tech: Python, InsightFace, Ultralytics, OpenCV
What I built: Final pipeline selection & eval, DB-backed embedding store, pricing analysis, reproducible scripts.

If you'd like broader work samples (ML, systems, course projects), I'm happy to share a tailored list on request.