search

ageitgey / face_recognition

The world's simplest facial recognition api for Python and the command line
MIT License
53.56k stars 13.5k forks source link

Realtime Face Recognition #1074

Open sushrutt12 opened 4 years ago

sushrutt12 commented 4 years ago

face_recognition version: 1.2.3 Python version: 3.6.5 Operating System: Windows 10 64bit GPU: NVidia GeForce MX150 10GB memory

Description

Frontend: I am trying to capture a frame via the webcam and send it to the backend. Backend: I use face_locations to find out the boxes for each face. Followed by face_encodings to generate the encodings. Then I run each face through my encodings to see if it is known. I then send the name and coordinates of the identified face to the front end and use a js library to mark the face. I want to achieve realtime face recognition, for this I have tried using the following two approaches:

In both cases it lags, I have already checked the performance of the js library this is not the reason for slowness, on running a profiler the bottleneck seems to be in face_locations & face_encodings. Any suggestions on how to make this work?

haris-ahmed commented 4 years ago

Instead of sending whole frame to backend for processing get faces from frame and send them to backend for recognition

try res10_300x300_ssd_iter_140000.caffemodel pretrained model for face detection https://www.pyimagesearch.com/2018/02/26/face-detection-with-opencv-and-deep-learning/

imZEH commented 4 years ago

@sushrutt12 How did you go in this area?. I've tried as well this kind of approach and it's lag.

sushrutt12 commented 4 years ago

@haris-ahmed Thanks but I did pursue it half way but I am working on something else now. @imZEH His approach did make things faster but for my use case my issue maybe the front end code.