Abstract:Conventional detectors suffer from performance degradation when dealing with long-tailed data due to a classification bias towards the majority head categories. In this paper, we contend that the learning bias originates from two factors: 1) the unequal competition arising from the imbalanced distribution of foreground categories, and 2) the lack of sample diversity in tail categories. To tackle these issues, we introduce a unified framework called BAlanced CLassification (BACL), which enables adaptive rectification of inequalities caused by disparities in category distribution and dynamic intensification of sample diversities in a synchronized manner. Specifically, a novel foreground classification balance loss (FCBL) is developed to ameliorate the domination of head categories and shift attention to difficult-to-differentiate categories by introducing pairwise class-aware margins and auto-adjusted weight terms, respectively. This loss prevents the over-suppression of tail categories in the context of unequal competition. Moreover, we propose a dynamic feature hallucination module (FHM), which enhances the representation of tail categories in the feature space by synthesizing hallucinated samples to introduce additional data variances. In this divide-and-conquer approach, BACL sets a new state-of-the-art on the challenging LVIS benchmark with a decoupled training pipeline, surpassing vanilla Faster R-CNN with ResNet-50-FPN by 5.8% AP and 16.1% AP for overall and tail categories. Extensive experiments demonstrate that BACL consistently achieves performance improvements across various datasets with different backbones and architectures. Code and models are available at <a class="link-external link-https" href="https://github.com/Tianhao-Qi/BACL" rel="external noopener nofollow">this https URL</a>.

The Devil is in Classification: A Simple Framework for Long-tail Object Detection and Instance Segmentation

On Model Calibration for Long-Tailed Object Detection and Instance Segmentation

Spatial Likelihood Voting with Self-Knowledge Distillation for Weakly Supervised Object Detection.

SLV: Spatial Likelihood Voting for Weakly Supervised Object Detection

DIM: Long-tailed Object Detection and Instance Segmentation via Dynamic Instance Memory

Seesaw Loss for Long-Tailed Instance Segmentation

Balanced Classification: A Unified Framework for Long-Tailed Object Detection

Long-tail Detection with Effective Class-Margins

Learning Box Regression and Mask Segmentation under Long-Tailed Distribution with Gradient Transfusing

Distance Metric-Based Learning for Long-Tail Object Detection

Adaptive Class Suppression Loss for Long-Tail Object Detection

Learning from Rich Semantics and Coarse Locations for Long-tailed Object Detection

The devil is in the object boundary: towards annotation-free instance segmentation using Foundation Models

Boosting Long-tailed Object Detection via Step-wise Learning on Smooth-tail Data

Fractal Calibration for long-tailed object detection

Long-tailed Visual Recognition with Deep Models: A Methodological Survey and Evaluation

Forest R-CNN: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentation

ForestDet: Large-Vocabulary Long-Tailed Object Detection and Instance Segmentation

SOLO: A Simple Framework for Instance Segmentation

Scalable Video Object Segmentation with Simplified Framework

Rectify the Regression Bias in Long-Tailed Object Detection