Equalization Loss for Long-Tailed Object Recognition

TL;DR

This paper introduces an equalization loss to improve long-tailed object recognition by preventing rare categories from being overwhelmed by gradients from common categories, leading to better detection performance.

Contribution

It proposes a novel equalization loss that selectively ignores gradients for rare categories, enhancing learning for long-tailed datasets like LVIS.

Findings

Achieved 4.1% AP gain on rare categories

Achieved 4.8% AP gain on common categories

Won 1st place in LVIS Challenge 2019

Abstract

Object recognition techniques using convolutional neural networks (CNN) have achieved great success. However, state-of-the-art object detection methods still perform poorly on large vocabulary and long-tailed datasets, e.g. LVIS. In this work, we analyze this problem from a novel perspective: each positive sample of one category can be seen as a negative sample for other categories, making the tail categories receive more discouraging gradients. Based on it, we propose a simple but effective loss, named equalization loss, to tackle the problem of long-tailed rare categories by simply ignoring those gradients for rare categories. The equalization loss protects the learning of rare categories from being at a disadvantage during the network parameter updating. Thus the model is capable of learning better discriminative features for objects of rare classes. Without any bells and whistles, our method achieves AP gains of 4.1% and 4.8% for the rare and common categories on the challenging LVIS benchmark, compared to the Mask R-CNN baseline. With the utilization of the effective equalization loss, we finally won the 1st place in the LVIS Challenge 2019. Code has been made available at: https: //github.com/tztztztztz/eql.detectron2