MFuseNet: Robust Depth Estimation with Learned Multiscopic Fusion
Yuan, Weihao, Fan, Rui, Wang, Michael Yu, Chen, Qifeng
–arXiv.org Artificial Intelligence
We design a multiscopic vision system that utilizes a low-cost monocular RGB camera to acquire accurate depth estimation. Unlike multi-view stereo with images captured at unconstrained camera poses, the proposed system controls the motion of a camera to capture a sequence of images in horizontally or vertically aligned positions with the same parallax. In this system, we propose a new heuristic method and a robust learning-based method to fuse multiple cost volumes between the reference image and its surrounding images. To obtain training data, we build a synthetic dataset with multiscopic images. The experiments on the real-world Middlebury dataset and real robot demonstration show that our multiscopic vision system outperforms traditional two-frame stereo matching methods in depth estimation. Our code and dataset are available at https://sites.google.com/view/multiscopic.
arXiv.org Artificial Intelligence
Aug-6-2021
- Genre:
- Research Report (0.82)
- Industry:
- Media
- Television (0.46)
- Photography (0.46)
- Film (0.46)
- Media
- Technology:
- Information Technology > Artificial Intelligence
- Vision > Image Understanding (1.00)
- Robots (1.00)
- Machine Learning > Neural Networks
- Deep Learning (0.46)
- Information Technology > Artificial Intelligence