Development an Object Tracking Method Based on Deep Learning

Object tracking based on deep learning is a challenging task. Accurate object detection and tracking requires a neural network that is deeper than a typical network. Therefore, it requires more levels of processing and takes more time. The YOLO model is considered one of the most state-of-the-art neural networks. Its wide application in object tracking depends on the speed of processing and extracting object features. In this article, we develop the YOLO model by changing the final layer from a fully connected layer to a fully convolutional layer, changing the activation function, and adding a batch normalization layer. A comparison of the results with the old YOLO model shows improved accuracy and speed. Additionally, the developed model is less sensitive to noisy environments.