Detect new objects better by teaching classifier not to ignore unlabeled objects
Detect new objects better by teaching classifier not to ignore unlabeled objects
Learning to Detect Every Thing in an Open World
arXiv paper abstract https://arxiv.org/abs/2112.01698
arXiv PDF paper https://arxiv.org/pdf/2112.01698.pdf
Project page https://ksaito-ut.github.io/openworld_ldet
Many … applications require the detection of novel objects, yet state-of-the-art object detection … do not excel at this task.
… their assumption … teaches the model to treat the unannotated objects as background.
… propose a simple yet surprisingly powerful data augmentation and training scheme we call Learning to Detect Every Thing (LDET).
… paste annotated objects on a background image sampled from a small region of the original image.
… training into two parts: 1) training … classification … head on augmented images, and 2) training the mask heads on original images.
… LDET … outperforming baselines on cross-category generalization on COCO, as well as cross-dataset evaluation on UVO and Cityscapes.
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
