The STONEFLY9 Image Database

• High resolution TIF images. gzipped tar file. 22.94 Gb.
• High resolution JPEG images. gzipped tar file. 2.2Gb.
• Low resolution JPEG images. gzipped tar file 526Mb.
• Extracted features. gzipped tar file. 355Mb. The features that are extracted include SIFT features computed using the following region detectors:
  • PCBR (Principal Curvature-Based Regions)
  • Kadir saliency
  • Hessian affine
  • Regular sampling along a grid
  In addition, we provide a shape representation based on edge descriptors and a segmentation mask for the bug in each image.

In all cases, there is one sub-folder for each taxon (cal, dor, hes, iso, mos, pte, swe, yor, zap). The information is further divided into three sets (set0, set1, set2) to support 3-fold cross-validation. In our experiments, one set was used for building dictionaries, one set for training the classifier, and one set for testing.

All uses of this data should cite the following papers:

• Lytle, D. A., Martínez-Muñoz, G., Zhang, W., Larios, N., Shapiro, L., Paasch, R., Moldenke, A., Mortensen, E. A., Todorovic, S., Dietterich, T. G. (2010). Automated processing and identification of benthic invertebrate samples. Journal of the North American Benthological Society, 29(3), 867-874. PDF preprint.

• Martínez-Muñoz, G., Zhang, W., Payet, N., Todorovic, S., Larios, N., Yamamuro, A., Lytle, D., Moldenke, A., Mortensen, E., Paasch, R., Shapiro, L., Dietterich, T. (2009). Dictionary-Free Categorization of Very Similar Objects via Stacked Evidence Trees.. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR-2009), Miami Beach, FL. PDF Preprint.