Depth Prediction from Images

alt text 

We tackle the problem of single image depth estimation, which, without additional knowledge, suffers from many ambiguities. We introduce a hierarchical representation of the scene and formulate single image depth estimation as inference in a graphical model whose edges let us encode the interactions within and across the different layers of our hierarchy. Our method therefore still produces detailed depth estimates, but also leverages higher-level information about the scene.

From Image to Concept

alt text 

We study basic-level categories for describing visual concepts, and empirically observe context-dependant basic-level names across thousands of concepts. We propose methods for predicting basic-level names using a series of classification and ranking tasks, producing the first large-scale catalogue of basic-level names for hundreds of thousands of images depicting thousands of visual concepts. We also demonstrate the usefulness of our method with a picture-to-word task.

Contour Detection and Completion

alt text 

  • Yansheng Ming, Hongdong Li, Xuming He, Winding Number for Region-Boundary Consistent Salient Contour Extraction, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013 [pdf]

  • Yansheng Ming, Hongdong Li, Xuming He, Connected Contours: a Contour Completion Model That Respects Closure-Effect, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2012 [pdf]

Image Understanding for Bionic Eye

alt text 

  • Xuming He, Junae Kim, Nick Barnes, An Face-based Visual Fixation System for Prosthetic Vision, Annual International Conference of the Engineering in Medicine and Biology Society (EMBC), 2012, USA

  • Tao Wang, Xuming He, Nick Barnes, Glass Object Localization by Joint Inference of Boundary and Depth, International Conference on Pattern Recognition (ICPR), 2012 [pdf] [Dataset]

Motion Anlaysis

alt text 

  • Shuang Wu, Xuming He, Hongjing Lu, and Alan Yuille, A Unified Model of Short-range and Long-range Motion Perception, Annual Conference on Neural Information Processing Systems (NIPS), 2010, Vancouver, Canada [pdf]

  • Xuming He and Alan Yuille, Occlusion Boundary Detection using Pseudo-Depth, European Conference on Computer Vision (ECCV), 2010, Greece [pdf]

Image/Scene Labeling

alt text 

  • Xuming He, and Richard S. Zemel, Latent Topic Random Fields: Learning Using a Taxonomy of Labels, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2008 [pdf (with Appendix)]

  • Xuming He, Richard Zemel, and Deb Ray, Learning and Incorporating Top-down Cues in Image Segmentation, European Conference on Computer Vision (ECCV), 2006. [pdf] [Dataset]

  • Xuming He, Richard Zemel, and Miguel Carreira-Perpinan, Multiscale Conditional Random Fields for Image Labelling, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2004 [pdf] [Dataset]