Nawaf-Rayhan585/Face-Recognition

Nawaf-Rayhan585/Face-Recognition has 41 GitHub stars and 13 forks. GitHot measured 0 stars in 24 hours. Last checked 2026-11-11 21:59 UTC. Last push 2026-09-27.

How these counts are measured

Lightweight face recognition template with Jarvis-style overlay — pure OpenCV (Haar + LBPH), runs on any laptop

41 stars 13 forks

Measured star growth

in 24h: 0
in 7 days: Collecting history
in 15 days: Collecting history
in 1 month: Collecting history

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

README.md

Original

Face Recognition Template 🧑‍🦱

Lightweight, reusable face recognition with a Jarvis-style dot overlay. Pure OpenCV — Haar cascade for detection, LBPH for recognition. No dlib, no cmake, no GPU, no model downloads. Runs on any laptop.

Works with: webcam · images · video files

Setup

pip install -r requirements.txt

That's it. opencv-contrib-python includes everything needed.

Register a person

python register_face.py John john.jpg

Use a clear front-facing photo. You can register the same person multiple times with different photos (different angles/lighting) — it improves accuracy, each call retrains the model automatically.

Run

python main.py                                   # webcam
python main.py --source photo.jpg --output result.jpg
python main.py --source video.mp4 --output result.mp4

Known face = name + cyan box + jarvis dots. Unknown = red box. Press q to quit (webcam/video mode).

Reuse for a new client

Copy this whole folder.
Delete everything inside data/faces/, plus data/lbph_model.yml and data/labels.txt.
Register new client's people.
Done — same code, new data.

Notes

CONFIDENCE_THRESHOLD in main.py (70) = strictness. LBPH confidence is a distance score — LOWER means better match, so lower threshold = stricter.
Accuracy improves a lot with 3-5 photos per person vs just 1.
Haar cascade is decent but not perfect at odd angles/low light — swap in a DNN detector later if a client needs higher accuracy.

Project structure

core/detector.py     # Haar detection + LBPH recognition
core/ui.py           # Jarvis-style dots + boxes
main.py              # run on webcam / image / video
register_face.py     # add a person
models/              # Haar cascade xml

Want higher accuracy?

This template is built to be light. For production (bad lighting, side angles, big crowds), swap the detector/recognizer for a deep model like InsightFace (ArcFace) or a YOLO face model. The rest of the code stays the same.

License

MIT — see LICENSE.

Built by Nawaf Rayhan · [email protected] · ⭐ the repo if it helped you