Places 205 Scene Dataset

This is a large-scale scene-centric image dataset with 205 common scene categories and 2.5 million images, mainly used for scene recognition.

The dataset includes:

• Training set: at least 5,000 images per category, at most 15,000 images, a total of 2,479,500 images
• Validation set: 100 images per class, 20,500 images in total
• Test set: 200 images per category, 41,000 images in total

Example image: