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 2104.02867cs.CVcs.AIcs.RO
收藏

基于示能迁移学习的人机交互检测

Affordance Transfer Learning for Human-Object Interaction Detection

The University of Sydney(悉尼大学);Shanghai AI Laboratory(上海人工智能实验室);Shenzhen Technology University(深圳技术大学)

Zhi Hou, Baosheng Yu, Yu Qiao, Xiaojiang Peng, Dacheng Tao

摘要:

推理人-物交互对于更深层次的场景理解是必不可少的,而物体功能性对于人类发现未知物体的人-物交互具有重要意义。受此启发,我们引入了一种启示迁移学习方法来联合检测HOI与新对象并识别启示。具体地,HOI表示可以被解耦为示能表示和对象表示的组合,使得可以通过组合示能表示和来自附加图像的新颖对象表示来组成新颖交互,即将示能表示转移到新颖对象。通过所提出的示能迁移学习,该模型还能够从已知的示能表示中推断新对象的示能。因此,所提出的方法可以用于1)改进HOI检测的性能,特别是对于具有不可见对象的HOI;以及2)推断新颖对象的示能表示。在两个数据集HICO-DET和HOI-COCO(来自V-COCO)上的实验结果表明,与最近用于HOI检测和对象示能表示检测的最先进方法相比,该方法有了显着的改进。代码可在https://github.com/zhihou7/HOI-CL上获得

英文摘要:

Reasoning the human-object interactions (HOI) is essential for deeper scene understanding, while object affordances (or functionalities) are of great importance for human to discover unseen HOIs with novel objects. Inspired by this, we introduce an affordance transfer learning approach to jointly detect HOIs with novel objects and recognize affordances. Specifically, HOI representations can be decoupled into a combination of affordance and object representations, making it possible to compose novel interactions by combining affordance representations and novel object representations from additional images, i.e. transferring the affordance to novel objects. With the proposed affordance transfer learning, the model is also capable of inferring the affordances of novel objects from known affordance representations. The proposed method can thus be used to 1) improve the performance of HOI detection, especially for the HOIs with unseen objects; and 2) infer the affordances of novel objects. Experimental results on two datasets, HICO-DET and HOI-COCO (from V-COCO), demonstrate significant improvements over recent state-of-the-art methods for HOI detection and object affordance detection. Code is available at https://github.com/zhihou7/HOI-CL

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。

↑