TL;DR: We propose STAR, a method that uses prototype replay to tackle distribution discrepancies between single-step training sets and the complete dataset in ...

5 Conclusion. In this paper, we present a novel class-incremental semantic segmentation (CISS) method named STAR that addresses the bias towards part of classes ...

This leads to an overrepresentation of these foreground classes in the single-step training set, causing the classification biased towards these classes.

Code release for "Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation" (NeurIPS ...

Existing class-incremental semantic segmentation (CISS) methods mainly tackle catastrophic forgetting and background shift, but often overlook another ...

Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation ... Segmentation. R Cong, H Xiong ...

STAR is a method for teaching computers to recognize and label objects in images. It helps the computer learn new objects without forgetting what it has ...

年份 · Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation. J Chen, R Cong, Y Luo ...

Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation ... Segmentation. R Cong, H Xiong ...

Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation [NeurIPS 2023] [paper] ...