DESCRIPTION
The dataset consists of 8 object categories (bike, shoe, car, iron, mouse, cellphone, stapler, toaster). For each object category, the dataset contains images of 10 individual object instances under 8 viewing angles, 3 heights and 3 scales for a total number of ∼7000 images. Images are roughly 400x300 pixels in bmp format.
An example of image filename is bicycle_02_A1_H3_S2.bmp
- "bicycle" indicates the category
- "02" indicates the object instance number
- "A1" indicates the viewing angle
- "H3" indicates the viewing height
- "S2" indicates the viewing distance (scale)
We have provided outlines the for each object in the dataset. The outlines are stored in subfolders indicated by mask/. A matlab code for extracting the outlines can be found below.
DOWNLOAD
HOW TO REFERENCE THIS DATASET
We would appreciate it if you cite our work when using the dataset:
S. Savarese and L. Fei-Fei, "3D generic object categorization, localization and pose estimation", IEEE Intern. Conf. in Computer Vision (ICCV), Brazil, October, 2007
-Last update: October 31, 2007
|