Due to the variations of viewpoint, pose, and illumination, a given individual may appear considerably different across different\ncamera views. Tracking individuals across camera networks with no overlapping fields is still a challenging problem. Previousworks\nmainly focus on feature representation and metric learning individually which tend to have a suboptimal solution. To address this\nissue, in this work, we propose a novel framework to do the feature representation learning and metric learning jointly. Different\nfrom previous works, we represent the pairs of pedestrian images as new resized input and use linear Support Vector Machine\nto replace softmax activation function for similarity learning. Particularly, dropout and data augmentation techniques are also\nemployed in this model to prevent the network from overfitting. Extensive experiments on two publically available datasets VIPeR\nand CUHK01 demonstrate the effectiveness of our proposed approach.
Loading....