IJCV-Data

Domain

• Evaluation of Localized Semantics in Images

Comments

• 1'014 images with manual segmentations and semantic labels for each segment and a methodology for using this kind of data for recognition evaluation.
  The evaluation methodology establishes protocols for mapping machine segmentation to human segmentation, scoring matches at different levels of specificity, and taking synonyms, sense ambiguity and multiple labels into accounted. Based on these protocols, were developed two evaluation approaches for measuring the range and the frequency of semantics that an algorithm can recognize correctly.

Media (image, video, mixed, …)

• Image

Size (no images, in GB, …)

Source (FlickR, Corel)

• The images and segmentations are from the UCB segmentation benchmark database (Martin et al., 2001).
  The database is extended by manually labeling each segment with its most specific semantic concept in WordNet

Annotation type (free text, structured, …)

Ground truth

Event or project

Task (retrieval, recognition, …)

Format

• images that were used for the labeling are in JPEG format
• the segmentation data for the above images are in the form of a text file with the extension .seg
• label data that has some meta information followed by the number of labels/segments and a list of the labels. They are in files have the extension .lab.

Quality (resolution)

Creation date

Copyright

• If any questions or comments, please email Quanfu Fan at quanfu AT cs DOT arizona DOT edu.

URL

• http://vision.cs.arizona.edu/quanfu/semantic/index.html