LPW stands for Labeled Pedestrian in the Wild, which is a pedestrian detection dataset. The dataset contains three different scenes and a total of 2,731 pedestrians, each of which is captured by 2 to 4 cameras. The salient features of LPW are: it contains 7,694 tracklets and more than 590,000 images.

LPW has three important differences from existing datasets: large-scale and clean, automatically detected bounding boxes, more crowded scenes, and a larger age span. This dataset provides a more realistic and challenging benchmark, which helps to further explore more powerful algorithms.