Deep_Learning_Face_Verification

Deep Learning Face Verification system with features like single shot multiple detect, real-time database update. Accuracy upto 97% (Hardware dependent).

In this system, we have integrated facial recognition algorithm with machine learning algorithm into the process of automatic attendance system. This system is implemented in basic and fundamental principle on the presence of a digital camera in the classroom. The digital camera would capture 2 images in the time interval of 25 minutes in a lecture of 50 minutes. Now image would be provided to system and system would extract all the faces from the image. Thereafter, the face would be compared with the existing trained model of faces and checks if face exists or not. If face exists on current database then the system would save unique ID of a student in attendance database or discards in case student doesn’t exist in classroom database. In this approach, we had also used Tensorflow estimator API to classify different faces using Deep Neural Network (DNN) which again gets trained from the images extracted in real time. The designed system does not interrupt class in any manner. Moreover, Feedback feature of the proposed project helps to train our model over verified images from the tested process. Therefore, it saves potential time of students as well as of teachers/administrators. Hence, increasing the accuracy over time. From the experiment analysis it is found that the accuracy of proposed system is 97%. Hence proposed system doesn’t require any rectification and verification from teachers.