.png){kind=icon}
.png){kind=icon}
返回
Task-balanced distillation for object detection
.png){kind=icon}
OA
AI
DOI：10.1016/j.patcog.2023.109320.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Mainstream object detectors are commonly constituted of two sub-tasks, including classification and re-gression tasks, implemented by two parallel heads. This classic design paradigm inevitably leads to in-consistent spatial distributions between classification score and localization quality (IOU). Therefore, this paper alleviates this misalignment in the view of knowledge distillation. First, we observe that the mas -sive teacher achieves a higher proportion of harmonious predictions than the lightweight student. Based on this intriguing observation, a novel Harmony Score (HS) is devised to estimate the alignment of clas-sification and regression qualities. HS models the relationship between two sub-tasks and is seen as prior knowledge to promote harmonious predictions for the student. Second, this spatial misalignment will result in inharmonious region selection when distilling features. To alleviate this problem, a novel Task-decoupled Feature Distillation (TFD) is proposed by flexibly balancing the contributions of classifica-tion and regression tasks. Eventually, HD and TFD constitute the proposed method, named Task-Balanced Distillation (TBD). Extensive experiments demonstrate the considerable potential and generalization of the proposed method. Notably, when equipped with TBD, the performances of RetinaNet-R18/RetinaNet-R50/Faster-RCNN-R18 can be boosted from 33.2/37.4/34.5 to 37.3/41.2/37.7, outperforming the recent KD-based methods like FRS, FGD, and MGD.(c) 2023 Elsevier Ltd. All rights reserved.
Keyword:
Object detection
Knowledge distillation
Computer vision
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。