Video Indexing through Human Faces by Combined Deep Learning Neural Networks

Abstract

This research aims to suggest an algorithm that uses the human face as a cue for detecting faces and recognition from input video. Face recognition has become popular because it has various applications, such as information security, smart cards, video surveillance, and law enforcement. The suggested approach, divided into two parts, combines the Multitask Convolution Neural Network (MTCNN) and Shuffle Net. Face detection from a video series using MTCNN and Shuffle Net is the first phase, and face recognition using eigenvalue recognition—a method for identifying faces using principal component analysis (PCA)—is the second. A Multitask convolution neural network with Shuffle Net and Eigen face recognition, a method for facial identification utilizing principal component analysis (PCA), is used to carry out these two processes, a deep learning-based tending type of neural network. Numerous experiments are run on various test datasets to assess the suggested strategy. The outcomes of the simulations are very intriguing and demonstrate how practical the suggested approach is. The human face plays a crucial role in applications like security systems, credit and debit card verification, and surveillance on identifying illegal public venues. So, face recognition is one of the most crucial techniques for video indexing. Facial recognition is becoming increasingly crucial in many aspects of our lives, such as security (discovering missing children speeds up searches for missing persons), attendance, healthcare, the retail sector, and banking. Detecting and recognizing faces are utilized for indexing after the human face has been identified in the input video. Using this video indexing method, we can quickly and effectively search for human activity in the input video. The suggested face detection approach is contrasted with the MTCNN and Shuffle Net algorithms for video indexing. After comparing, it is found that using the combined MTCNN and Shuffle Net algorithm for face detection is more effective and time-saving than just MTCNN and Shuffle net. Using the combined MTCNN and Shuffle Net method, more faces are discovered. Among other advantages, Eigen face recognition using PCA is straightforward, efficient, and precise. It can work with low-resolution photographs and adjust to variations in lighting, facial expressions, and head tilt. However, it has some limitations, such as its sensitivity to changes in face size and shape and its inability to handle partial faces, occlusions, or disguises.CNN also uses facial recognition, but the volume of images is where they confront their biggest obstacle. CNN needs many training images to attain a high level of recognition accuracy. Eigen facial recognition using PCA produces good results on fewer training images. For the indicated strategy, it is 99.35%. Human faces in the input video are the results of face detection and identification, and videos are indexed using these faces as cues.