Project link: http://lsun.cs.princeton.edu/2017/

LSUN challenge provides a large scale scene understanding dataset covering three major problems:

• Scene classification
• Segmentation task on street images
• Saliency prediction

In scene classification problems, the expected output of the algorithm is the top most likely scene category in the image. At the time of writing, there are 10 different classes such as bedroom, classroom, and restaurant. In the segmentation problem, you can try to solve the pixel-level segmentation and instance-specific segmentation. In saliency prediction problems, the goal is to predict where a human looks in a scene image.