arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2301.01019cs.CV
收藏

Correlation Loss:加强分类和本地化之间的相关性

Correlation Loss: Enforcing Correlation between Classification and Localization

Middle East Technical University (METU)(中东技术大学)

Fehmi Kahraman, Kemal Oksuz, Sinan Kalkan, Emre Akbas

摘要:

对象检测器通常通过分类和定位损失的加权和来训练。最近的研究(例如,用辅助头预测IoU,广义焦点损失,秩和排序损失)已经表明,迫使这两个损失项以非常规的方式彼此相互作用产生了有用的归纳偏差并提高了性能。受这些工作的启发,我们专注于分类和定位之间的相关性,并作出了两个主要贡献:(i)我们提供了一个分析的目标检测器中的分类和定位任务之间的相关性的影响。我们确定为什么相关性会影响各种基于NMS和无NMS的检测器的性能,我们设计的措施来评估相关性的影响,并使用它们来分析常见的检测器。(ii)根据我们的观察,例如,无NMS检测器也可以从相关中受益,我们提出了相关损失,一种新的插入式损失函数,通过直接优化相关系数来提高各种对象检测器的性能:例如,Sparse R-CNN的相关性损失是一种无NMS方法,在COCO上产生1.6 AP增益,在Cityscapes数据集上产生1.8 AP增益。我们在Sparse R-CNN上的最佳模型达到51.0 AP,没有COCO test-dev上的测试时间增强,达到了最先进的水平。代码可在https://github.com/fehmikahraman/CorrLoss上获得

英文摘要:

Object detectors are conventionally trained by a weighted sum of classification and localization losses. Recent studies (e.g., predicting IoU with an auxiliary head, Generalized Focal Loss, Rank & Sort Loss) have shown that forcing these two loss terms to interact with each other in non-conventional ways creates a useful inductive bias and improves performance. Inspired by these works, we focus on the correlation between classification and localization and make two main contributions: (i) We provide an analysis about the effects of correlation between classification and localization tasks in object detectors. We identify why correlation affects the performance of various NMS-based and NMS-free detectors, and we devise measures to evaluate the effect of correlation and use them to analyze common detectors. (ii) Motivated by our observations, e.g., that NMS-free detectors can also benefit from correlation, we propose Correlation Loss, a novel plug-in loss function that improves the performance of various object detectors by directly optimizing correlation coefficients: E.g., Correlation Loss on Sparse R-CNN, an NMS-free method, yields 1.6 AP gain on COCO and 1.8 AP gain on Cityscapes dataset. Our best model on Sparse R-CNN reaches 51.0 AP without test-time augmentation on COCO test-dev, reaching state-of-the-art. Code is available at https://github.com/fehmikahraman/CorrLoss

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑