Dynamic Scale Training for Object Detection

TL;DR

The paper introduces Dynamic Scale Training (DST), a feedback-guided data preparation method that significantly improves object detection performance across scales without extra inference costs.

Contribution

DST is a novel, simple paradigm that uses optimization feedback to dynamically guide scale data preparation, outperforming traditional multi-scale strategies.

Findings

Achieves over 2% AP improvement on MS COCO

Generalizes well across different backbones and tasks

Enables faster convergence during training

Abstract

We propose a Dynamic Scale Training paradigm (abbreviated as DST) to mitigate scale variation challenge in object detection. Previous strategies like image pyramid, multi-scale training, and their variants are aiming at preparing scale-invariant data for model optimization. However, the preparation procedure is unaware of the following optimization process that restricts their capability in handling the scale variation. Instead, in our paradigm, we use feedback information from the optimization process to dynamically guide the data preparation. The proposed method is surprisingly simple yet obtains significant gains (2%+ Average Precision on MS COCO dataset), outperforming previous methods. Experimental results demonstrate the efficacy of our proposed DST method towards scale variation handling. It could also generalize to various backbones, benchmarks, and other challenging downstream tasks like instance segmentation. It does not introduce inference overhead and could serve as a free lunch for general detection configurations. Besides, it also facilitates efficient training due to fast convergence. Code and models are available at github.com/yukang2017/Stitcher.