Bootstrap Your Object Detector via Mixed Training

TL;DR

MixTraining is a novel training paradigm for object detection that leverages mixed augmentations and pseudo boxes, improving detector performance by addressing augmentation strength and annotation errors through bootstrapping.

Contribution

It introduces MixTraining, a new method that enhances object detection training by combining augmentation strategies and pseudo labels via detector bootstrapping.

Findings

Consistent performance improvements on COCO dataset.

Faster R-CNN with ResNet-50 improved from 41.7 to 44.0 mAP.

Cascade-RCNN with Swin-Small improved from 50.9 to 52.8 mAP.

Abstract

We introduce MixTraining, a new training paradigm for object detection that can improve the performance of existing detectors for free. MixTraining enhances data augmentation by utilizing augmentations of different strengths while excluding the strong augmentations of certain training samples that may be detrimental to training. In addition, it addresses localization noise and missing labels in human annotations by incorporating pseudo boxes that can compensate for these errors. Both of these MixTraining capabilities are made possible through bootstrapping on the detector, which can be used to predict the difficulty of training on a strong augmentation, as well as to generate reliable pseudo boxes thanks to the robustness of neural networks to labeling error. MixTraining is found to bring consistent improvements across various detectors on the COCO dataset. In particular, the performance of Faster R-CNN \cite{ren2015faster} with a ResNet-50 \cite{he2016deep} backbone is improved from 41.7 mAP to 44.0 mAP, and the accuracy of Cascade-RCNN \cite{cai2018cascade} with a Swin-Small \cite{liu2021swin} backbone is raised from 50.9 mAP to 52.8 mAP. The code and models will be made publicly available at \url{https://github.com/MendelXu/MixTraining}.