Occlusion detector using convolutional neural network for person re-identification

Abstract

Technique of comparing pedestrian images observed by different cameras to determine whether they are the same person is important in the surveillance system. This technique is called Person re-identification. Most of Person reidentification is underway assuming that occlusion does not occur. However, since occlusion occurs frequently in the surveillance system and affects accuracy, it is necessary to determine whether the occlusion occurs before applying person re-identification in the real environment. In order to deal with occlusion, we introduce occlusion detector based convolutional neural networks that determine occlusion of an input image. We also created an occlusion dataset through data augmentation and learned the occlusion detector using this dataset. We have achieved 98.7% accuracy of the data obtained by synthesizing occlusion in public dataset.