Long-tailed Instance Segmentation using Gumbel Optimized Loss

TL;DR

This paper introduces Gumbel Optimized Loss (GOL), a novel loss function designed for long-tailed instance segmentation that improves detection and segmentation of rare classes by aligning with the Gumbel distribution, outperforming existing methods.

Contribution

The paper proposes GOL, a new loss function tailored for long-tailed detection and segmentation, addressing limitations of sigmoid and softmax functions in imbalanced datasets.

Findings

GOL outperforms state-of-the-art by 1.1% AP on LVIS.

Segmentation accuracy improves by 9.0%.

Detection of rare classes increases by 20.3%.

Abstract

Major advancements have been made in the field of object detection and segmentation recently. However, when it comes to rare categories, the state-of-the-art methods fail to detect them, resulting in a significant performance gap between rare and frequent categories. In this paper, we identify that Sigmoid or Softmax functions used in deep detectors are a major reason for low performance and are sub-optimal for long-tailed detection and segmentation. To address this, we develop a Gumbel Optimized Loss (GOL), for long-tailed detection and segmentation. It aligns with the Gumbel distribution of rare classes in imbalanced datasets, considering the fact that most classes in long-tailed detection have low expected probability. The proposed GOL significantly outperforms the best state-of-the-art method by 1.1% on AP , and boosts the overall segmentation by 9.0% and detection by 8.0%, particularly improving detection of rare classes by 20.3%, compared to Mask-RCNN, on LVIS dataset. Code available at: https://github.com/kostas1515/GOL