![]() ![]() In the training phase of CNN, we randomly choose a rectangle region in the image, and erase its pixels with random values. 1 Introduction Figure 1: Examples of Random Erasing. Random Erasing is parameter learning free, easy to implement, and can be integrated into most of the CNN-based recognition models.Īlbeit simple, Random Erasing yields consistent improvement in image classification, object detection and person re-identification (re-ID).įor image classification, our method improves WRN-28-10: top-1 error rate from 3.72% to 3.08% on CIFAR10, and from 18.68% to 17.65% on CIFAR100.įor object detection on PASCAL VOC 2007, Random Erasing improves Fast-RCNN from 74.8% to 76.2% in mAP.įor person re-ID, when using Random Erasing in recent deep models, we achieve the state-of-the-art accuracy: the rank-1 accuracy is 89.13% for Market-1501, 84.02% for DukeMTMC-reID, and 63.93% for CUHK03 under the new evaluation protocol. ![]() In this process, training images with various levels of occlusion are generated, which reduce the risk of network overfitting and make the model robust to occlusion. ![]() In training phase, Random Erasing randomly selects a rectangle region in an image, and erases its pixels with random values. In this paper, we introduce Random Erasing, a simple yet effective data augmentation techniques for training the convolutional neural network (CNN). ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |