Deep View: A Dynamite Wireless Facial Recognition System With Data Logging

Traditional facial recognition techniques use static algorithms that are non-adaptive to changes in pose, expression, and other input image factors. This paper proposed a dynamic wireless facial recognition system with data logging capabilities called the deep view that used a convolutional neural network (CNN). The Deep views approach was divided into face detection and face recognition. For face detection, a Histogram of Oriented Gradients (HOG)-based technique was used in conjunction with Face Alignment through Affine Transformation for input image preprocessing. The facial recognition stage utilizes an open-face implementation for the neural network, modified clustering for grouping identities, and Destiny-Based Spatial Clustering of Applications with Noise (DBSCAN) for removing outliers. The system is camera invariant such that recognition is retained from camera to camera regardless of specification. It yielded an accuracy of 87.03% and the processing time of each face was found to be 13.7 ms so a 10 fps framerate was employed. Data logging was implemented by creating an archive of the time, date, camera number, and picture of the encounter for each distinct identity. The images are stored in a separate folder sorted by date. This information can be exported as a spreadsheet file. Also, a face-searching functionality was added to the system. This enables the user to upload an external photo and search the database for a matching identity. Successful results were obtained for the face recognition system and data logging but deep view demonstrated no superhuman skills. Generally, the system can still be used in real-work scenarios.