DETECTION OF ACCURATE FACIAL DETECTION USING HYBRID DEEP CONVOLUTIONAL RECURRENT NEURAL NETWORK

Abstract
Facial Landmark discovery is an imperative issue in numerous PC vision applications about appearances. It is extremely testing as human faces in wild conditions regularly present expansive varieties fit as a fiddle because of various stances, impediments or demeanors. Profound neural systems have been connected to take in the guide from face pictures to confront shapes. To the best of our insight, Recurrent Neural Network (RNN) has not been utilized in this issue yet. In this paper, we propose a technique which uses RNN and Deep Neural Network (DNN) to take in the face shape. To start with, we design a system utilizing Convolutional Neural Network (CNN) to get the underlying Landmark estimation of appearances. At that point, we utilize feed-forward neural systems for neighborhood look where a segment based seeking technique is investigated. By utilizing LSTMCNN-RNN, the underlying estimation is more dependable which makes the accompanying segment based pursuit doable and exact. Tests demonstrate that the worldwide system utilizing CNN-LSTM-RNN shows signs of improvement results than past systems in the two recordings and single picture. Our technique beats the cutting edge calculations particularly regarding fine estimation of Landmark spots.

Authors
M Sivaram, V Porkodi, Amin Salih Mohammed, V Manikandan
Lebanese French University, Iraqi Kurdistan

Keywords
Facial landmark, Deep Neural Network, Recurrent Neural Network, Convolutional Neural Network
Published By :
ICTACT
Published In :
ICTACT Journal on Soft Computing
( Volume: 9 , Issue: 2 )
Date of Publication :
Januray 2019
DOI :

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.