CNNとDPを用いたトラッキング

抄録

This study examines a way of tracking using a Convolutional Neural Network (CNN), which is used extensively in the computer science field, and Dynamic Programming (DP) for getting the optimal tracking path. In this study, we use the CNN score of the object for tracking through the learning of the Cifar10 dataset. We set the inputs of CNN to all image patches which each frame of video is divided into. After getting the CNN score of each image patch in all frames of video, we get the optimal tracking path by DP.