04 - Face Recognition - II

Assignment

  1. Refer to this beautiful blog (Links to an external site.).
  2. Collect 10 facial images of 10 people you know (stars, politicians, etc). The more the images you collect, the better your experience would be. Add it to this LFW (Links to an external site.) dataset.
  3. Train as in the blog and upload the FR model to Lambda
  4. Share the link to “that” single page.
  5. Share the link to GitHub repo.

Solution

160x160 Face Aligned dataset:
https://github.com/satyajitghana/TSAI-DeepVision-EVA4.0-Phase-2/tree/master/04-FaceRecognition-II/indian_face_dataset_160

Dataset Creation: https://github.com/satyajitghana/TSAI-DeepVision-EVA4.0-Phase-2/blob/master/04-FaceRecognition-II/CustomLFW%26Indian_Face_Dataset.ipynb

[The above dataset was then added to LFW Dataset]

LFW-Plus-Dataset (LFW + 10 Indian People): https://drive.google.com/file/d/17k_3guDwSKclVjpyxEmgSmRvB0Z64LCq/view?usp=sharing

Training:
(Training on LFW + 10 Indian People)
https://github.com/satyajitghana/TSAI-DeepVision-EVA4.0-Phase-2/blob/master/04-FaceRecognition-II/Train_MTCNN_LFW_Plus.ipynb

(Training on 10 faces only)
https://github.com/satyajitghana/TSAI-DeepVision-EVA4.0-Phase-2/blob/master/04-FaceRecognition-II/Train_MTCNN_Indian_Face.ipynb

All of the TensorClan Models can be found at https://drive.google.com/drive/folders/15y4WyvaJI9AsfioKUNX1inJtrDqn2UvV?usp=sharing

Dataset

Indian Faces Dataset
enter image description here

LFW Dataset
enter image description here

Backend: https://thetensorclan-backend.herokuapp.com/

Code for FrontEnd: https://github.com/extensive-vision-ai/thetensorclan-web
Code for BackEnd: https://github.com/extensive-vision-ai/thetensorclan-backend-heroku

Demo

LFW - Plus Recognizer (contains 10 additional people)

enter image description here

Indian Faces Recognizer

enter image description here

[right click -> open image in a new tab] for better experience

Train Metrics

enter image description here

Explanation

Remember that we had created FaceAlign previously in Face-Recognition-I ? we simply supply our face image to that AWS Lambda API and get the aligned face, now we have added the face recognition model to heroku(why ? see notes), and now its simply matter of sending the aligned face to the model backend and fetching the results ! that’s it !

Notes

Personal Notes