Description

The Mall dataset (Loy, Chen, Gong, Xiang) is a video-based crowd-counting benchmark captured from a single fixed surveillance camera in a shopping mall. It contains 2000 frames with per-pedestrian annotations and is one of the standard low-to-medium-density crowd-counting datasets, alongside UCSD. Together with UCSD it is regularly criticised in the literature for low scene-perspective variation, since both come from a single continuous video sequence.

Modality / size

  • Modality: RGB video frames from a fixed surveillance camera.
  • Subjects / scenarios: shopping mall, single viewpoint.
  • Labels: head annotations per frame, 2000 frames.

Used by (papers)

  • Standard CNN crowd-counting baseline alongside ShanghaiTech, UCF_CC_50, WorldExpo '10, and UCSD.
  • Used as the video baseline in vault papers comparing CSI / RF crowd counting against camera-based counting.

Notes

  • Public dataset; mirror at the original Loy et al. project page.
  • Disambiguation: Mall extracted from claims is noisy because the literal word "mall" matches many unrelated phrases; only treat as the dataset when context is crowd counting.

1 vault paper evaluate on this dataset

Titles and DOIs only — no abstracts, no analyses.

  • A survey of recent advances in CNN-based single image crowd counting and density estimation 2018 DOI ↗