Few-shot Semantic Segmentation with Self-supervision from Pseudo-classes
This addresses the challenge of generalization in few-shot semantic segmentation for computer vision applications, but it is incremental as it builds on existing state-of-the-art methods.
The paper tackled the problem of few-shot semantic segmentation, where existing methods perform poorly when query images contain multiple semantic classes, by proposing a self-supervised task that generates pseudo-classes in the background to provide extra training data. The result was an improvement in meanIoU by 2.5% to 6.7% on benchmarks like PASCAL-5i and COCO.
Despite the success of deep learning methods for semantic segmentation, few-shot semantic segmentation remains a challenging task due to the limited training data and the generalisation requirement for unseen classes. While recent progress has been particularly encouraging, we discover that existing methods tend to have poor performance in terms of meanIoU when query images contain other semantic classes besides the target class. To address this issue, we propose a novel self-supervised task that generates random pseudo-classes in the background of the query images, providing extra training data that would otherwise be unavailable when predicting individual target classes. To that end, we adopted superpixel segmentation for generating the pseudo-classes. With this extra supervision, we improved the meanIoU performance of the state-of-the-art method by 2.5% and 5.1% on the one-shot tasks, as well as 6.7% and 4.4% on the five-shot tasks, on the PASCAL-5i and COCO benchmarks, respectively.