People counting: Datasets & Models

Manually counting people from CCTV monitors is tedious, error-prone, and impossible to scale across dozens of cameras or long recording windows. Security and surveillance teams need reliable occupancy numbers for compliance limits, queue management, and incident response, and human operators simply cannot watch every feed continuously. Automating people counting detection with a vision model turns raw camera footage into structured, timestamped counts without adding staff. The result is consistent, 24/7 monitoring that flags overcrowding or unusual footfall the moment it happens.

Automated people counting delivers continuous, repeatable headcounts that integrate directly with the IP cameras and NVR infrastructure you already operate, so there is no need to rip and replace hardware. Detection-based counting scales from a single doorway to entire multi-site deployments, and models like RF-DETR balance accuracy on dense, overlapping crowds with the throughput needed for real-time streams. Teams gain auditable occupancy data, faster response to crowding events, and a foundation for downstream analytics like dwell time and flow direction.

1. Upload an image: Drop any photo into the widget above to try people counting for free, right on this page.
2. Preview instant results: The model detects person, head, face, backpack, helmet within seconds and returns an annotated image with structured predictions, so you can immediately see how well it works. Adjust the detection classes and re-run until the results fit your scenario.
3. Like what you see?: With a free Roboflow account, you can deploy this workflow with the Serverless Cloud API, customize it in your workspace, or train RF-DETR on your own images for production accuracy.
4. Get started for free: No code or training data required to begin: upload, review the results, and make it part of your stack in a single click.