arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1608.08021cs.CV
收藏

PVANET:用于实时的目标检测的深度轻量级神经网络

PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection

Intel Imaging and Camera Technology(英特尔成像与相机技术)

Kye-Hyeon Kim, Sanghoon Hong, Byungseok Roh, Yeongjae Cheon, Minje Park

摘要:

本文介绍了我们如何在多类别目标检测任务中实现最先进的准确性,同时通过适应和结合最近的技术创新,最大限度地减少计算成本。遵循“CNN特征提取+区域建议+ RoI分类”的通用流水线,我们主要重新设计了特征提取部分,因为区域建议部分计算量不大,分类部分可以使用截断SVD等常用技术有效压缩。我们的设计原则是“更少的通道,更多的层”,并采用一些构建块,包括级联ReLU,Inception和HyperNet。所设计的网络是深和薄的,并训练的帮助下,批量归一化,剩余连接,和学习率调度的平台检测。我们在著名的目标检测基准测试中获得了可靠的结果:在VOC 2007上获得了83.8%的mAP(平均精度),在VOC 2012上获得了82.5%的mAP(第二名),而在Intel i7- 6700 K CPU上仅获得了750 ms/图像,在NVIDIA Titan X GPU上仅获得了46 ms/图像。从理论上讲,我们的网络只需要12.3%的计算成本相比,ResNet-101,在VOC 2012的赢家。

英文摘要:

This paper presents how we can achieve the state-of-the-art accuracy in multi-category object detection task while minimizing the computational cost by adapting and combining recent technical innovations. Following the common pipeline of "CNN feature extraction + region proposal + RoI classification", we mainly redesign the feature extraction part, since region proposal part is not computationally expensive and classification part can be efficiently compressed with common techniques like truncated SVD. Our design principle is "less channels with more layers" and adoption of some building blocks including concatenated ReLU, Inception, and HyperNet. The designed network is deep and thin and trained with the help of batch normalization, residual connections, and learning rate scheduling based on plateau detection. We obtained solid results on well-known object detection benchmarks: 83.8% mAP (mean average precision) on VOC2007 and 82.5% mAP on VOC2012 (2nd place), while taking only 750ms/image on Intel i7-6700K CPU with a single core and 46ms/image on NVIDIA Titan X GPU. Theoretically, our network requires only 12.3% of the computational cost compared to ResNet-101, the winner on VOC2012.

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。

↑