Pattern Macthing with Optimization using Machine Learning and Deep Learning | Programming portfolio by Yash Patel

here the inference image has different facial orientation and Obscured face, even few people will fail to recognize this image of his struggling days but still our model accurately recognized the face

We did not trained virat kohli’s image, and similarity score obtained by our model exceeds our threshold value set, therefore it gives result as unknown person

- RHS image is the image which is saved by the user for later recognition and whose embeddings are stored in .pickle file - Trained Image of Salman was without Beard, image we given to system to be recognized is having beard, still our system recognises person accurately

- User upload the images with a label which needs to be recognized later by the system - All the images are processed, the face is detected by the MTCNN Algorithm and detected face image will be saved in the database - For each image(face) 128-d vector representation is generated by a pre-trained dlib model or facenet model. - These encodings are then dumped in a .pickle file for future inference purposes. - For inference, again encodings are generated for face present in the input image. - Using a threshold value as a strictness measure, Euclidean distances are compared through the pickle file and the most closest face is recognized to be that of the user. - Name and ID of the recognized user are displayed.

Source:github.com/yashpatel7025/ml_dl_project-pattern_matching_with_optimization

Published:June 28, 2021