Better object detection with few examples by using image rather than proposals with Meta-DETR
Better object detection with few examples by using image rather than proposals with Meta-DETR
Meta-DETR: Image-Level Few-Shot Detection with Inter-Class Correlation Exploitation
arXiv paper abstract https://arxiv.org/abs/2208.00219v1
arXiv PDF paper https://arxiv.org/pdf/2208.00219v1.pdf
GitHub https://github.com/ZhangGongjie/Meta-DETR
Few-shot object detection has been extensively investigated by incorporating meta-learning into region-based detection frameworks.
… paradigm … still constrained by … (i) low-quality region proposals for novel classes and (ii) negligence of the inter-class correlation among different classes.
… design Meta-DETR, which (i) is the first image-level few-shot detector, and (ii) introduces a novel inter-class correlational meta-learning strategy to capture and leverage the correlation among different classes
… Meta-DETR works entirely at image level without any region proposals, which circumvents the constraint of inaccurate proposals
… Meta-DETR … simultaneously attend to multiple support classes within a single feedforward, which allows to capture the inter-class correlation among different classes, thus significantly reducing the misclassification … and enhancing … generalization to novel classes.
… show that the proposed Meta-DETR outperforms state-of-the-art methods by large margins …
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
