/Real-time-Human-Authentication

Smart India Hackathon 2020 DRDO problem statement: CK107. Multiple camera-based dynamic human authentication for a gate. Our solution involves multiple cameras and we authenticate humans using face, gait, pose, etc features.

Primary LanguageJavaScript

The_Eminents_SIH20

DRDO problem statement (CK107) : Video based dynamic human authentication system for access control

Multiple camera based dynamic human authentication for a gate. Our solution involves multiple cameras and we authenticate humans using face, gait, pose etc features. Please refer the system architecture below.

Architecture

ARCHITECTURE

Working System Videos:

  1. Main system with dashboard : Check from 0:48 timestamp
  2. Anti Piggy Backing using head detection
  3. Robust Face Recognition : Multiple light conditions
  4. Robust Face Detection : Crowd recognition
  5. Automated Video saving for future retrieval

Benchmarking

We produced robust benchmaring of various state-of-the-art models. Check out the benchmarking folder for details.

Presentations:

  1. Mentoring Round 1
  2. Evaluation 1
  3. Mentoring Round 2
  4. Evaluation 2

Python Library for Face Recognition, Detection and Landmarking

Face_Lib

Whole code for project along with deep learning models

Final Code and Models