3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions
Project Webpage: http://3dmatch.cs.princeton.edu/
Abstract:
Matching local geometric features on real-world depth images is a challenging task due to the noisy, low-resolution, and incomplete nature of 3D scan data. These difficulties limit the performance of current state-of-art methods, which are typically based on histograms over geometric properties. In this paper, we present 3DMatch, a data-driven model that learns a local volumetric patch descriptor for establishing correspondences between partial 3D data. To amass training data for our model, we propose an unsupervised feature learning method that leverages the millions of correspondence labels found in existing RGB-D reconstructions. Experiments show that our descriptor is not only able to match local geometry in new scenes for reconstruction, but also generalize to different tasks and spatial scales (e.g. instance-level object model alignment for the Amazon Picking Challenge, and mesh surface correspondence). Results show that 3DMatch consistently outperforms other state-of-the-art approaches by a significant margin.
Видео 3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions канала Andy Zeng
Abstract:
Matching local geometric features on real-world depth images is a challenging task due to the noisy, low-resolution, and incomplete nature of 3D scan data. These difficulties limit the performance of current state-of-art methods, which are typically based on histograms over geometric properties. In this paper, we present 3DMatch, a data-driven model that learns a local volumetric patch descriptor for establishing correspondences between partial 3D data. To amass training data for our model, we propose an unsupervised feature learning method that leverages the millions of correspondence labels found in existing RGB-D reconstructions. Experiments show that our descriptor is not only able to match local geometry in new scenes for reconstruction, but also generalize to different tasks and spatial scales (e.g. instance-level object model alignment for the Amazon Picking Challenge, and mesh surface correspondence). Results show that 3DMatch consistently outperforms other state-of-the-art approaches by a significant margin.
Видео 3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions канала Andy Zeng
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Multi-view Self-supervised Deep Learning for 6D Pose Estimation in the Amazon Picking Challenge](https://i.ytimg.com/vi/BjL-DW4jVEM/default.jpg)
![Learning Synergies between Pushing and Grasping with Self-supervised Deep Reinforcement Learning](https://i.ytimg.com/vi/-OkyX7ZlhiU/default.jpg)
![VNect: Real-time 3D Human Pose Estimation with a Single RGB Camera - SIGGRAPH2017](https://i.ytimg.com/vi/W1ZNFfftx2E/default.jpg)
![Transporter Networks: Rearranging the Visual World for Robotic Manipulation](https://i.ytimg.com/vi/496UVuAdOP4/default.jpg)
![TossingBot: Learning to Throw Arbitrary Objects with Residual Physics](https://i.ytimg.com/vi/f5Zn2Up2RjQ/default.jpg)
![Robots Learning to Toss](https://i.ytimg.com/vi/-O-E1nFm6-A/default.jpg)
![5 VERY COOL open source machine learning projects](https://i.ytimg.com/vi/zbaXiXrYCyo/default.jpg)
![정밀한 측정시대 3D 스캐닝 / YTN 사이언스](https://i.ytimg.com/vi/qKaGvN1K0uI/default.jpg)
![ML Engine - Machine Learning in the Cloud](https://i.ytimg.com/vi/fsv0rty7QhU/default.jpg)
![Detection of free parking spaces v0.1](https://i.ytimg.com/vi/B6p98QVEPGA/default.jpg)
![Top 10 Machine Learning And Artificial Intelligence Projects | Engineering Projects](https://i.ytimg.com/vi/X69SOQrF5jQ/default.jpg)
![anagliph video 3d](https://i.ytimg.com/vi/YW-ml-GKrb0/default.jpg)