Computer Vision - ECCV 2016 Workshops - Amsterdam, The Netherlands, October 8-10 and 15-16, 2016, Proceedings, Part I
Computer Vision - ECCV 2016 Workshops - Amsterdam, The Netherlands, October 8-10 and 15-16, 2016, Proceedings, Part I
复制标题
计算机视觉 - ECCV 2016 研讨会 - 荷兰阿姆斯特丹，2016 年 10 月 8-10 日和 15-16 日，会议记录，第一部分
DOI：
10.1007/978-3-319-46604-0_57
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Westlake N
中科院分区：
文献类型：
--
作者：
Westlake N
CNN大大提高了照片中物体检测的性能。然而，对艺术品中目标检测的研究仍然有限。我们在一个具有挑战性的数据集People-Art上展示了最先进的性能，该数据集包含来自照片，卡通和41种不同艺术作品运动的人物。我们通过为这项任务微调CNN来实现这种高性能，从而也证明了在照片上训练CNN会导致照片的过度拟合：只有前三层或前四层从照片转移到艺术品。虽然CNN的表现是最高的，但它仍然低于60%的AP，这表明需要进一步的工作来解决交叉描述问题。
CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset,People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by fine-tuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN’s performance is the highest yet, it remains less than 60 % AP, suggesting further work is needed for the cross-depiction problem.