Abstract
Publications home
Estimating low-rank region likelihood maps
at
| Gabriela Csurka, Zoltan Kato, Andor Juhasz, Martin Humenberger |
| Computer Vision and Pattern Recognition (CVPR 2020), Washington, United States, 16-18 June, 2020 |
| Download |
A planar region in an image is low-rank if it is possible to recover a canonical view which rectifies the region. This typically holds for repetitive pattern (corners, edges, etc.) on planar surfaces. Such a region enables obtaining the camera orientation/position with respect to a 3D plane using the TILT (Transform Invariant Low-rank Textures) algorithm. This is even possible without using any prior knowledge of the visual information (features) on that plane. However, there is no method which automatically detects such regions and a brute force solution applied on all image patches is not feasible because of the heavy computation workload (iterative convex optimization) needed.
To overcome this limitation, we introduce a self-supervised region proposal network to directly extract planar regions containing a low-rank texture from an image. First, we use TILT to compute a novel low-rank likelihood map using a sliding window approach at multiple scales and at given steps. Second, we use these maps to train our network to generate such maps directly from new images. We evaluate or method on real-world datasets and show the performance by comparing it with an implementation of the current state-of-the-art, the hand-crafted method TILT. The results show that our method reliably detects low-rank regions which can successfully be used to compute accurate camera orientations/positions.
To make robots autonomous in real-world everyday spaces, they should be able to learn from their interactions within these spaces, how to best execute tasks specified by non-expert users in a safe and reliable way. To do so requires sequential decision-making skills that combine machine learning, adaptive planning and control in uncertain environments as well as solving hard combinatorial optimisation problems. Our research combines expertise in reinforcement learning, computer vision, robotic control, sim2real transfer, large multimodal foundation models and neural combinatorial optimisation to build AI-based architectures and algorithms to improve robot autonomy and robustness when completing everyday complex tasks in constantly changing environments.
VISION
The research we conduct on expressive visual representations is applicable to visual search, object detection, image classification and the automatic extraction of 3D human poses and shapes that can be used for human behavior understanding and prediction, human-robot interaction or even avatar animation. We also extract 3D information from images that can be used for intelligent robot navigation, augmented reality and the 3D reconstruction of objects, buildings or even entire cities.
Our work covers the spectrum from unsupervised to supervised approaches, and from very deep architectures to very compact ones. We’re excited about the promise of big data to bring big performance gains to our algorithms but also passionate about the challenge of working in data-scarce and low-power scenarios.
Furthermore, we believe that a modern computer vision system needs to be able to continuously adapt itself to its environment and to improve itself via lifelong learning. Our driving goal is to use our research to deliver embodied intelligence to our users in robotics, autonomous driving, via phone cameras and any other visual means to reach people wherever they may be.
This web site uses cookies for the site search, to display videos and for aggregate site analytics.
Learn more about these cookies in our privacy notice.
Cookie settings
You may choose which kind of cookies you allow when visiting this website. Click on "Save cookie settings" to apply your choice.
This content is currently blocked. To view the content please either 'Accept social media cookies' or 'Accept all cookies'.
For more information on cookies see our privacy notice.