The usage of computer vision in identifying pedestrians attributes has received a great attention, especially in the visual surveillance systems. For instance, searching for system based on the attributes. Attributes Identification using Convolutional Neural Network architecture is presented in this article, since the architecture can perform feature learning. CNN consist of convolution layer, ReLU, Pooling, and Fully-connected. There are three experiment scenarios are conducted based on the number of convolution layers, to determine the effect of layers on CNN performance. Three different CNN architectures were trained and tested using a PETA dataset with 35 attributes. The highest accuracy achieved is 75.66% based on number of convolutional layers. The conducted experiments showed that more numbers of convolution layers used would produce the better CNN's performance.
Copyrights © 2020