The objective of this project of is to build an elementary interface to apply modern algorithm of Face Recognition on real-time video from Webcam. The first implementation is based on:
- MTCNN (Multi-task Cascaded Convolutional Networks) to faces detection
- and inception_resnet trained on VGGFace2 to extract face features and recognition
This can identify myself (Simon):
Me in my room while covid-19 confinement (the kids crying in the living room).
🚧 Next algorithms as SDD (Single Shot Detector ) and Yolo (You only look once) will be added later. :construction:
My contribution is an interface with the details of MTCNN and the possibility to custom the network:
- Number of boxes after each neuronal networks
- Possibility to switch from cpu to gpu
- Update the thresholds at each levels
The program is organized according to the following diagram:
