DATA SETS

App-Icons Data-Set

5 batches that include 1401 photos and 501 ground truth masks, 13MB

This data-set includes several batches:

  • withIcons.zip
    501 snapshots of icons, taken with different iPhones by dozens of users. A typical snapshot will include an icon occupying a big part of the image, and its surrounding which is the iPhone’s background, or a website background.
  • GTwithIcons.zip
    Ground Truth mask of “withIcons” batch.
  • withoutIcons.zip
    399 snapshots taken with different iPhones by dozens of users, that do not include a whole icon within.
  • testIcons.zip
    251 photos of icons. These photos were automatically cropped from snapshots of icons. Hence, a typical photo is fully occupied by an icon.
  • trainIcons.zip
    250 photos of icons, and the original icons of the corresponding app in App-Store. The original icons are located in the OriginalIcons folder. The photos are organized in folders, such that each folder contains several photos of the same original icon.

AppsIcons* We are sorry, we are not allowed to release the entire App-Store icons database. In order to obtain that, please address Apple Inc. For example through their EPF program.

These batches were used in the following paper, reading it will clarify why they were organized in such way.

I. Friedman, L. Zelnik-Manor, “Icon Scanning: Towards Next Generation QR Codes”, Computer Vision and Pattern Recognition, Providence, Rhode-Island, USA, Jun. 2012.

Archeological

11 objects, 66 MB

Archeological DataSetThis page includes several archaeological artifacts that were found in the excavations of Tel Dor city, Israel. These models are protected by copyrights, but maybe used for academic research purposes. Commercial usage is prohibited.

If you use these objects for your research, please add acknowledgements to the Laboratory of Computer Graphics & Multimedia at the Technion and to the Zinman Institute of Archaeology, University of Haifa.

Download database.

Sea Shells

4 objects, 23 MB

Sea Shells DataSet This page includes several SeaShells models that were scanned in our lab. These models are protected by copyrights, but maybe used for academic research purposes. Commercial usage is prohibited.

If you use these objects for your research, please add acknowledgements to the Laboratory of Computer Graphics & Multimedia at the Technion.

Download database.

Pinecones

5 objects, 54 MB

Pinecones DataSetThis page includes several Pinecons models that were scanned in our lab. These models are protected by copyrights, but maybe used for academic research purposes. Commercial usage is prohibited.

If you use these objects for your research, please add acknowledgements to the Laboratory of Computer Graphics & Multimedia at the Technion.

Download database.

Horns

3 objects, 17 MB

Horns DataSetThis page includes several Horns models that were scanned in our lab. These models are protected by copyrights, but maybe used for academic research purposes. Commercial usage is prohibited.

If you use these objects for your research, please add acknowledgements to the Laboratory of Computer Graphics & Multimedia at the Technion.

Download database.

YouTube Categories

Our data-set contains 11 video categories. All videos are in spatial resolution of 320 x 240

To download a category click on the corresponding icon.

If you use this data please cite the following work:
S. Zanetti, L. Zelnik-Manor, P. Perona, “A walk through the web’s video clips”, The First IEEE Workshop on Internet Vision, in Computer Vision and Pattern Recognition (CVPR08), Anchorage, Alaska, Jun. 2008

Categories:

  • Autos & Vehicles
  • Comedy
  • Entertainment
  • Film & Animation
  • Gadgets & Games
  • Howto & DIY
  • Music
  • News & Politics
  • Pets & Animals
  • Sports
  • Travel & Places

Multi-View Actions

This data-set extends the IXMAS-actions data-set

It includes videos of 8 different people performing 13 different actions filmed by 8 different cameras. Each person repeated the entire set of actions 3 times. We further provide background masks, which are not very accurate. If you obtain better ones, and you’re willing to share, please let us know. Please refer to the README file for further details.The original database can be found here

If you use this data please cite the following work:
D. Rudoy and L. Zelnik-Manor, “Viewpoint Selection for Human Actions”, International Journal of Computer Vision (IJCV), 2011

Categories:

  • Videos
  • Setup
  • Ground-truth
  • Background Masks