LFSD, the full name of which is Light Field Saliency Database, is a light field dataset for saliency detection.

The dataset contains 100 light fields with a spatial resolution of 360×360 pixels. Each light field data includes 5 types of images, namely, 5-dimensional undecoded original light field file, focus stack, full focus image, depth map, and salient target truth map. The scene usually contains only one salient target located in the center of the foreground and is large in size.