News:
- The presentation on feature evaluation given at the Workshop in Miami is now available: Feature Evaluation.
- Executable feature detectors for linux, Windows and Mac OS that include Harris-..., Hessian-..., DoG, MSER, SIFT, ColorSIFT will be made available soon.
About:
In the last decade we have seen enormous progress in the area of invariant local image and video descriptors that led to substantial improvements in many computer vision areas including registration, stereo vision, motion estimation, matching, retrieval, recognition of objects and actions. Techniques for low level image or video description are of wide interest in the computer vision and pattern recognition communities as they are the first stage of many algorithms. The purpose of this webpage is to encourage researchers to present their latest developments in local feature extraction for the general use or for specific applications. There are still many challenges left for a variety of problem classes. In particular, there is a need for diverse measurements from image and
video, robust to large changes of viewing conditions, extreme
photometric variations, occlusion, background clutter as well as
intra-class variations and specific applications i.e., category level correspondence.. We hope this webpage will serve to establish a standard evaluation protocol and test data for local features.
Performance evaluation:
This webpage provides an independent evaluation of local features. The test is focused on recognition/categorization related to PASCAL VOC challenge as well as matching of image sets with variations in viewing and acquisition conditions. Please see evaluation page for more details.
