MNIST: handwritten digits: The most commonly used sanity check. Dataset of 25x25, centered, B&W handwritten digits. It is an easy taskjust because something works on MNIST, doesnt mean it works.
Pictures of objects belonging to 256 categoriesPictures of objects belonging to 256 categories.
Binocular images of toy figurines under various illumination and pose.
32x32 color images with 10 / 100 categories. Not commonly used anymore, though once again, can be an interesting sanity check.
Vector data for the entire planet under a free license. It contains (an older version of) the US Census Bureaus data.
Different objects imaged at every angle in a 360 rotation.
COIL100 : Different objects imaged at every angle in a 360 rotation.
House numbers from Google Street View. Think of this as recurrent MNIST in the wild.
The CALTECH 101 dataset by Li Fei-Fei contains images for 101 categories with about 40 to 800 images per category. Most categories have about 50 images ...
A large dataset of annotated images.
The de-facto image dataset for new algorithms. Many image API companies have labels from their REST interfaces that are suspiciously close to the 1000 c...
Satellite shots of the entire Earth surface, updated every several weeks.
Generic image Segmentation / classificationnot terribly useful for building real-world image annotation, but great for baselines
is an image recognition dataset for developing unsupervised feature learning, deep learning, self-taught learning algorithms. Like CIFAR-10 with some mo...
Doppler radar scans of atmospheric conditions in the US.
Generic image understanding / captioning, with an associated competition.
A collection of 9 million URLs to images that have been annotated with labels spanning over 6,000 categories under Creative Commons.
Scene understanding with many ancillary tasks (room layout estimation, saliency prediction, etc.) and an associated competition.