Dense Depth Posterior (DDP) From Single Image and Sparse Range. Yang, Y., Wong, A., & Soatto, S. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3348–3357, June, 2019. ISSN: 2575-7075
doi  abstract   bibtex   
We present a deep learning system to infer the posterior distribution of a dense depth map associated with an image, by exploiting sparse range measurements, for instance from a lidar. While the lidar may provide a depth value for a small percentage of the pixels, we exploit regularities reflected in the training set to complete the map so as to have a probability over depth for each pixel in the image. We exploit a Conditional Prior Network, that allows associating a probability to each depth value given an image, and combine it with a likelihood term that uses the sparse measurements. Optionally we can also exploit the availability of stereo during training, but in any case only require a single image and a sparse point cloud at run-time. We test our approach on both unsupervised and supervised depth completion using the KITTI benchmark, and improve the state-of-the-art in both.
@inproceedings{yang_dense_2019,
	title = {Dense {Depth} {Posterior} ({DDP}) {From} {Single} {Image} and {Sparse} {Range}},
	doi = {10.1109/CVPR.2019.00347},
	abstract = {We present a deep learning system to infer the posterior distribution of a dense depth map associated with an image, by exploiting sparse range measurements, for instance from a lidar. While the lidar may provide a depth value for a small percentage of the pixels, we exploit regularities reflected in the training set to complete the map so as to have a probability over depth for each pixel in the image. We exploit a Conditional Prior Network, that allows associating a probability to each depth value given an image, and combine it with a likelihood term that uses the sparse measurements. Optionally we can also exploit the availability of stereo during training, but in any case only require a single image and a sparse point cloud at run-time. We test our approach on both unsupervised and supervised depth completion using the KITTI benchmark, and improve the state-of-the-art in both.},
	language = {en},
	booktitle = {2019 {IEEE}/{CVF} {Conference} on {Computer} {Vision} and {Pattern} {Recognition} ({CVPR})},
	author = {Yang, Yanchao and Wong, Alex and Soatto, Stefano},
	month = jun,
	year = {2019},
	note = {ISSN: 2575-7075},
	keywords = {\#CVPR{\textgreater}19, \#PointCloud, \#Sparse, /unread, 3D from Multiview and Sensors, 3D from Single Image, Benchmark testing, Codes, Computer vision, Deep learning, Laser radar, Point cloud compression, Robotics + Driving, Scene Analysis and Understanding, Training, ⭐⭐⭐⭐⭐},
	pages = {3348--3357},
}

Downloads: 0