Project link: http://image-net.org/

ImageNet is a large scale visual recognition challenge that has runs annually since 2010. The dataset is organized according to the WorkNet hierarchy. There are over ten million URLs of images with hand-annotated labels to indicate what objects are in the picture. There are at least one million images that have bounding boxes included.

The ImageNet challenge is held every year to evaluate algorithms for the following three problems:

• Object localization for 1,000 categories.
• Object detection for 200 fully-labeled categories.
• Object detection from video for 30 fully labeled categories. In July 17, 2017, the results of the 2017 challenge were announced with many advanced and interesting algorithms.