Interclass Prototype Relation for Few-Shot Segmentation
Atsuro Okazawa
Unverified — Be the first to reproduce this paper.
ReproduceAbstract
Traditional semantic segmentation requires a large labeled image dataset and can only be predicted within predefined classes. To solve this problem, few-shot segmentation, which requires only a handful of annotations for the new target class, is important. However, with few-shot segmentation, the target class data distribution in the feature space is sparse and has low coverage because of the slight variations in the sample data. Setting the classification boundary that properly separates the target class from other classes is an impossible task. In particular, it is difficult to classify classes that are similar to the target class near the boundary. This study proposes the Interclass Prototype Relation Network (IPRNet), which improves the separation performance by reducing the similarity between other classes. We conducted extensive experiments with Pascal-5i and COCO-20i and showed that IPRNet provides the best segmentation performance compared with previous research.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| COCO-20i (1-shot) | IPRNet (ResNet-50) | Mean IoU | 45.3 | — | Unverified |
| COCO-20i (1-shot) | IPRNet (ResNet-101) | Mean IoU | 46.9 | — | Unverified |
| COCO-20i (5-shot) | IPRNet (ResNet-101) | Mean IoU | 53.3 | — | Unverified |
| COCO-20i (5-shot) | IPRNet (ResNet-50) | Mean IoU | 51.1 | — | Unverified |
| PASCAL-5i (1-Shot) | IPRNet (ResNet-50) | Mean IoU | 65.7 | — | Unverified |
| PASCAL-5i (1-Shot) | IPRNet (ResNet-101) | Mean IoU | 67.5 | — | Unverified |
| PASCAL-5i (5-Shot) | IPRNet (ResNet-101) | Mean IoU | 70.9 | — | Unverified |
| PASCAL-5i (5-Shot) | IPRNet (ResNet-50) | Mean IoU | 70.2 | — | Unverified |