|
 |
Resources |
- General reference material on computer vision:
- Software
- Specific references on recognition:
- ICCV 2005 and CVPR 2007 short courses on recognizing and learning object categories.
- Data Sets and Tools
- The MNIST database of handwritten digits, from New York University. Tens of thousands of examples from zip codes, preprocessed for image normalization, labelled.
- The CMU database of frontal faces. Faces are from real-life pictures, with manual annotation of the face positions. Only the download of the entire database works.
- The COIL-100 database from Columbia University. Individual, carefully lit objects positioned on a neutral background.
- The NORB database from New York University. Fifty white-painted toys from five categories are viewed under controlled and systematically varied lighting and viewpoint conditions.
- The Caltech 101 database with 101 object categories, 40+ real-life images per category. Objects are manually outlined.
- The Caltech 256 database. Like Caltech 101, but with more categories, more images, and more challenging viewing conditions.
- The LabelMe database is being augmented daily by web users who volunteer to outline and label objects in images. The current, annotated database can be downloaded, including a Matlab toolbox.
- A collection of 80 million tiny images, downloaded by Antonio Torralba, Rob Fergus, and William T. Freeman through image search engines. Images are associated with the words used in the query that retrieved each of them.
- SnapMyLife, a picture-sharing site with labels attached to most images. Labels are entered either by hand or by an otherwise unspecified "automatic service." Using images and labels from this site would require writing a web crawler.
|
| |
|
|
|