The Caltech 256 is considered an improvement to its predecessor, the Caltech 101 dataset, with new features such as larger category sizes, new and larger clutter categories, and overall increased difficulty. This is a great dataset to train models for visual recognition: How can we recognize frogs, cell phones, sail boats and many other categories in cluttered pictures? How can we learn these categories in the first place? Can we endow machines with the same ability?

Content

There are 30,607 images in this dataset spanning 257 object categories. Object categories are extremely diverse, ranging from grasshopper to tuning fork. The distribution of images per category are:

• Min: 80

• Med: 100

• Mean: 119

• Max: 827

Acknowledgements

Original data source and banner image: http://www.vision.caltech.edu/Image_Datasets/Caltech256/

When using this dataset, please remember to cite:

Griffin, G. Holub, AD. Perona, P.
The Caltech 256.
Caltech Technical Report.