Embedded Files
Clothing-ADC, an image dataset containing one million images with over 1,000 subclasses for each clothing type. Our dataset offers a rich hierarchy of categories, creating well-defined sub-populations that support research on a variety of complex and novel tasks. To further facilitate the exploration of the aforementioned challenges (label noise detection and learning, class-imbalanced learning), we customize three benchmark subsets and provide benchmark performances of the implemented methods in our software. This offers researchers a platform for performance comparisons, enhancing the evaluation and refinement of their approaches.
3D scans of one person in 71 different poses, with correspondence. Email davis at soe.ucsc.edu for password access to the data page linked above. No commercial use. For information about the data see Stanford page (2005)
3D scans of 3000 different people, with correspondence. This correspondence also aligns with the SCAPE data above. (2014)
More than 100 objects under 90 different illumination conditions. We used this for relighting research. It might also be useful for lighting invariant object recognition, or photometric stereo research. (2011)
A scanned face in motion with 120 frames, individual meshes per frame. From SpaceTime Stereo paper. (2005)
Page updated
Report abuse