arXiv 2411.028612024-11-05cs.CV

基于中心度的实例感知知识蒸馏与任务级相互提升用于无人机图像目标检测

Centerness-based Instance-aware Knowledge Distillation with Task-wise Mutual Lifting for Object Detection on Drone Imagery

Beihang University（北京航空航天大学）；State Key Laboratory of Software Development Environment（软件开发环境国家重点实验室）；School of Computer Science and Engineering, Beihang University（北京航空航天大学计算机科学与工程学院）；Hangzhou Innovation Institute, Beihang University（北京航空航天大学杭州创新研究院）

Bowei Du, Zhixuan Liao, Yanan Zhang, Zhi Cai, Jiaxin Chen, Di Huang

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文首次将知识蒸馏应用于无人机图像目标检测，提出基于中心度的实例感知蒸馏（CID）策略和任务级轻量相互提升（Light-ML）模块，解决低前景背景比和小实例复杂背景问题，在VisDrone、UAVDT和COCO基准上提升了现有最先进KD方法的精度。

摘要：

开发用于无人机图像的准确且高效的检测器具有挑战性，因为空中场景具有固有的复杂性。虽然一些现有方法通过利用更大的模型来实现高精度，但其计算成本对于无人机来说过高。近年来，知识蒸馏（KD）在通用目标检测中展现出在显著压缩模型的同时保持令人满意的精度的巨大潜力。考虑到KD的优势，本文首次尝试将其应用于无人机图像的目标检测，并解决了两个内在问题：（1）低前景-背景比率和（2）小实例与复杂背景，这些问题导致训练不足，从而造成蒸馏不充分。因此，我们提出了一种基于中心度的实例感知蒸馏（CID）策略，以及一个任务级轻量相互提升（Light-ML）模块。Light-ML模块通过通道混洗和卷积相互协调分类和定位分支，在反向传播期间整合跨不同任务的教师监督，从而促进学生模型的训练。CID策略通过提议的中心度提取实例周围的有价值区域，增强蒸馏效果。在VisDrone、UAVDT和COCO基准上的实验表明，所提出的方法在计算需求相当的情况下提升了现有最先进KD方法的精度。代码将在录用后提供。

英文摘要：

Developing accurate and efficient detectors for drone imagery is challenging due to the inherent complexity of aerial scenes. While some existing methods aim to achieve high accuracy by utilizing larger models, their computational cost is prohibitive for drones. Recently, Knowledge Distillation (KD) has shown promising potential for maintaining satisfactory accuracy while significantly compressing models in general object detection. Considering the advantages of KD, this paper presents the first attempt to adapt it to object detection on drone imagery and addresses two intrinsic issues: (1) low foreground-background ratio and (2) small instances and complex backgrounds, which lead to inadequate training, resulting insufficient distillation. Therefore, we propose a task-wise Lightweight Mutual Lifting (Light-ML) module with a Centerness-based Instance-aware Distillation (CID) strategy. The Light-ML module mutually harmonizes the classification and localization branches by channel shuffling and convolution, integrating teacher supervision across different tasks during back-propagation, thus facilitating training the student model. The CID strategy extracts valuable regions surrounding instances through the centerness of proposals, enhancing distillation efficacy. Experiments on the VisDrone, UAVDT, and COCO benchmarks demonstrate that the proposed approach promotes the accuracies of existing state-of-the-art KD methods with comparable computational requirements. Codes will be available upon acceptance.

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。