Inspired by this finding, we present Stitcher, a feedback-driven data provider, which aims to train object detectors in a balanced way. In Stitcher, images are ...

In this paper, we present Stitcher, a feedback-driven data provider, which aims to train object detectors in a balanced way. In Stitcher, images are resized ...

Stitcher is presented, a feedback-driven data provider, which aims to train object detectors in a balanced way, and steadily improves performance by a large ...

Stitcher proposed two methods to stitch multiple images into one, either along spatial dimension or along channel dimension. The first one is similar to the ...

Apr 26, 2020 · A Simple and Effective Use of Object-Centric Images for Long-Tailed Object Detection. Object frequencies in daily scenes follow a long-tailed ...

Apr 26, 2020 · We propose a Dynamic Scale Training paradigm (abbreviated as DST) to mitigate scale variation challenge in object detection. Previous strategies ...

Apr 28, 2020 · Object detectors commonly vary quality according to scales, where the performance on small objects is the least satisfying.

Dec 19, 2021 · 算法流程如下图所示,如果第t次迭代的损失值中,小目标的贡献率低于阈值,则第t+1次迭代采用Stitcher,否则仍然采用常规图片.

Apr 29, 2020 · 在本文中，我们研究了这种现象，并发现：在大多数训练迭代中，小目标的损失对总损失几乎没有贡献，导致优化不平衡导致性能下降。受此启发，我们提出 ...