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

Source: [Boosting Long-tailed Object Detection via Step-wise Learning on Smooth-tail Data](<https://www.sota2.com/research/paper/68fa6f651e209131cea45496>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T01:19:28.160Z

Authors: Na Dong, Yongqiang Zhang, Mingli Ding, Gim Hee Lee

Published: 2023-05-22T08:53:50.000Z

## Abstract

Real-world data tends to follow a long-tailed distribution, where the class imbalance results in dominance of the head classes during training. In this paper, we propose a frustratingly simple but effective step-wise learning framework to gradually enhance the capability of the model in detecting all categories of long-tailed datasets. Specifically, we build smooth-tail data where the long-tailed distribution of categories decays smoothly to correct the bias towards head classes. We pre-train a model on the whole long-tailed data to preserve discriminability between all categories. We then fine-tune the class-agnostic modules of the pre-trained model on the head class dominant replay data to get a head class expert model with improved decision boundaries from all categories. Finally, we train a unified model on the tail class dominant replay data while transferring knowledge from the head class expert model to ensure accurate detection of all categories. Extensive experiments on long-tailed datasets LVIS v0.5 and LVIS v1.0 demonstrate the superior performance of our method, where we can improve the AP with ResNet-50 backbone from 27.0% to 30.3% AP, and especially for the rare categories from 15.5% to 24.9% AP. Our best model using ResNet-101 backbone can achieve 30.7% AP, which suppresses all existing detectors using the same backbone.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Object Detection on LVIS v1.0 (val)](<https://www.sota2.com/research/sota/object-detection-on-lvis-v1-0-val.md>): Object Detection; LVIS v1.0 (val); APbbox: 28.7.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2305.12833>)

[PDF](<https://arxiv.org/pdf/2305.12833.pdf>)
