Automatic face recognition and finding occurrence of actors in movies

Abstract

In this paper, artist face recognition, and all movie prediction system is proposed. It comprises of two phases. Initially, the faces in the video are recognized using an l1-minimization CNN + HOG framework, and some keyframes are selected, based on a robust measure of confidence. Then the labels are propagated from the keyframes to the remaining frames, by using transductive learning. The constraints in both feature and temporal spaces are integrated simultaneously. The output of the algorithm is tested on Indian Movie Face—Database and generated all the movies of those actors/actresses of the film.