Exploring Classification Equilibrium in Long-Tailed Object Detection

TLDR

Conventional detectors suffer imbalanced classification and performance drops when training data are severely skewed. This work proposes to use the mean classification score as an indicator of per‑category accuracy during training. We introduce an Equilibrium Loss that adjusts decision boundaries with a score‑guided margin for weak classes, and a Memory‑augmented Feature Sampling that oversamples weak‑class features, jointly achieving classification equilibrium and improving tail‑class performance while preserving head‑class accuracy, as evaluated on LVIS with Mask R‑CNN backbones. The method raises tail‑class AP by 15.6 points and surpasses recent long‑tailed detectors by over 1 AP. Code is available at https://github.com/fcjian/LOCE.

Abstract

The conventional detectors tend to make imbalanced classification and suffer performance drop, when the distribution of the training data is severely skewed. In this paper, we propose to use the mean classification score to indicate the classification accuracy for each category during training. Based on this indicator, we balance the classification via an Equilibrium Loss (EBL) and a Memory-augmented Feature Sampling (MFS) method. Specifically, EBL increases the intensity of the adjustment of the decision boundary for the weak classes by a designed score-guided loss margin between any two classes. On the other hand, MFS improves the frequency and accuracy of the adjustment of the decision boundary for the weak classes through over-sampling the instance features of those classes. Therefore, EBL and MFS work collaboratively for finding the classification equilibrium in long-tailed detection, and dramatically improve the performance of tail classes while maintaining or even improving the performance of head classes. We conduct experiments on LVIS using Mask R-CNN with various backbones including ResNet-50-FPN and ResNet-101-FPN to show the superiority of the proposed method. It improves the detection performance of tail classes by 15.6 AP, and outperforms the most recent long-tailed object detectors by more than 1 AP. Code is available at https://github.com/fcjian/LOCE.