Project Title: Face identification in a video sequence.
This project is related to face recognition and identification area. We have to search for a person in a video. In the face detection and identification we have to train our machine for thousands of faces, but for the purpose of searching, we have only one image of the person we are looking for. So we have to use face comparison system like faceNet system. We will start with noise reduction in video. Face detection will be next step and Face comparison on detected faces will be the final step. It will work on the stored videos and one input image with face to search.
The noise in frames and pixel difference between two consecutive frames have to reduce for better results. Noise can be reduced using Gaussian noise reduction method.
In case of video, we will have all angled faces. So we can’t use viola Jones algorithm for face detection. We can use CNN for face detection.
For face comparison, we will use FaceNet system. FaceNet system was developed by Google and it uses 128-D model for faces. 128-D model is nothing but the measurement of the faces in different directions and angles.
- Face Detection
- Face recognition
- Face identification
- Gaussian noise reduction method
- Convolution Neural Network
- FaceNet system
Tools Used: Python, openCV and FaceNet library.
1. Fraud detection in passports and visa
2. Identification of criminals
3. ATM and bank video surveillance
4. Prevent fraud voters
5. Find a thief or terrorist in stored video database from surveillance.
Note: This project has been started and face detection and recognition are initiated on frontal faces only.
GitHub link: https://github.com/kapilggg10/project1