Elankeerthana R. and Vinotha R. |
Dept.of Information Tech., M.Kumarasamy College of Engineering, Karur, Tamil Nadu, India.
Abstract
Convolutional network in deep learning algorithm has many architectural structures for the person re-identification to increase the network accuracy. Observation of the same person can be matched in different occasions like time, cameras is the Person recognition on appearance based classification. Many years researchers on computer vision realized Person re-identification has been tricky problem which video includes frame taken at different place, pose, condition, lighting condition, camera, occlusions, background and appearance. Here we implement different architectural structure in convolutional network in our dataset to give different accuracy and different error rate to analyses the best structural to recognize the person in different cameras. Tracking and finding a person deals with the security issue where many places like airports, streets, colleges, shopping malls, theaters and many other public places for identification of fraud cases.
Keywords: Convolutional neural network, Deep learning, Architectural Structure (LeNet, AlexNet, VGGNet, GoogLeNet, ResNet, ResNext, DenseNet), Person Recognition.
View PDF