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 2203.10747cs.CV
收藏

EAutoDet:一种高效的目标检测体系结构搜索算法

EAutoDet: Efficient Architecture Search for Object Detection

Shanghai Jiao Tong University(上海交通大学);Guangdong OPPO Mobile Telecommunications Co., Ltd.(广东欧珀移动通信有限公司)

Xiaoxing Wang, Jiale Lin, Junchi Yan, Juanping Zhao, Xiaokang Yang

摘要:

训练CNN进行检测是耗时的,因为数据集很大,网络模块很复杂,很难直接在检测数据集上搜索架构,这通常需要大量的搜索成本(通常是几十甚至几百个GPU天)。相比之下,本文介绍了一个高效的框架,名为EAutoDet,可以发现实用的骨干和FPN架构的目标检测在1.4 GPU天。具体来说,我们构造了一个超网的骨干和FPN模块,并采用可微的方法。为了减少GPU的内存需求和计算成本,我们提出了一种内核重用技术,通过共享一个边缘上的候选操作的权重,并将它们合并到一个卷积。还引入了一种动态信道细化策略来搜索信道号。大量的实验表明,我们的方法显着的功效和效率。特别是,所发现的架构超越了最先进的目标检测NAS方法,并在COCO测试开发集上实现了40.1 mAP(120 FPS)和49.2 mAP(41.3 FPS)。我们还将发现的结构转移到旋转检测任务中,在DOTA-v1. 0测试集上使用21. 1M参数实现了77. 05 mAP $_{\text{50}}$。

英文摘要:

Training CNN for detection is time-consuming due to the large dataset and complex network modules, making it hard to search architectures on detection datasets directly, which usually requires vast search costs (usually tens and even hundreds of GPU-days). In contrast, this paper introduces an efficient framework, named EAutoDet, that can discover practical backbone and FPN architectures for object detection in 1.4 GPU-days. Specifically, we construct a supernet for both backbone and FPN modules and adopt the differentiable method. To reduce the GPU memory requirement and computational cost, we propose a kernel reusing technique by sharing the weights of candidate operations on one edge and consolidating them into one convolution. A dynamic channel refinement strategy is also introduced to search channel numbers. Extensive experiments show significant efficacy and efficiency of our method. In particular, the discovered architectures surpass state-of-the-art object detection NAS methods and achieve 40.1 mAP with 120 FPS and 49.2 mAP with 41.3 FPS on COCO test-dev set. We also transfer the discovered architectures to rotation detection task, which achieve 77.05 mAP$_{\text{50}}$ on DOTA-v1.0 test set with 21.1M parameters.

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。

↑