人物专区最新影视资讯精选1230日论文推荐
Paper 研习社每日精选论文推荐 1230
Scalable Fine-grained Generated Image Classification Based on Deep Measurement Learning
作者：Xuan Xinsheng /Peng Bo /Wang Wei /Dong Jing
发表时间：2019/12/10
论文链接：https://paper.yanxishe.com/review/7899
推荐理由：这篇文章提出了一个基于深度度量学习的可扩展多类别分类框架，旨在更好地对生成图像进行分类。该框架增加了可伸缩性，以应对不断涌现的新型生成图像，并通过微调使模型对新型生成数据获得更好的检测性能。
Merging Weak and Active Supervision for Semantic Parsing
作者：Ni Ansong /Yin Pengcheng /Neubig Graham
发表时间：2019/11/29
论文链接：https://paper.yanxishe.com/review/7900
推荐理由：这篇文章研究了弱监督与主动学习之间的交集，以改进在弱监督下训练的语析器。作者评估了用于选择示例进行查询的各种主动学习启发式方法，以及针对此类查询的各种形式额外监督。在两个不同的数据集上评估了其方法，结果显示该方法可以提高弱监督解析器的性能。
Detecting GAN generated errors
Authors: Zhu Xiru, Che Fengdi, Yang Tianzi, Yu Tzuyang, Meger David, Dudek Gregory
Date of publication: 2019-12-02
Paper link: https://paper.yanxishe.com/review/7901
Recommendation reason: This paper proposes a novel method to detect errors in generated images by comparing actual images with the generated ones. The authors calculate the belongingness of each pixel to either the real distribution or the generated distribution using Wasserstein distance and attention mechanism.
A Billion Ways to Grasp: An Evaluation of Grasp Sampling Schemes>
Author(s): Eppner Clemens; Mousavian Arsalan; Fox Dieter.
Date published (online): December 11th, 2019.
Paper link (URL): https://paper.yanxishe.com/review/7902.
Recommendation reason: This paper reviews and compares different grasp sampling strategies based on SE(3) fine-grained discretization and physical simulations evaluating parallel jaw grasping quality and robustness. The authors evaluate YCB dataset's objects' more than billion possible grasps.
Look, Read and Feel: Benchmarking Ads Understanding with Multimodal Multitask Learning
