Crowd Estimation
Nodeflux Crowd Estimation is a VisionAI analytics software which aims to measure the number of people at large in a crowd. While the common practice is to count each person automatically like People Density analytics, which requires more time to consume to get an insightful report, this analytic will provide you the estimated number of people in a particular area of interest.
Camera Specification
| Camera Settings | Face Recognition | License Plate Recognition | General Analytics |
|---|---|---|---|
Frame Resolution | 1080p - 4K | 1080p - 4K | 720p - 1080p |
Codec | H.264 | H.264 | H.264 |
Codec Profile | High | Main | Baseline |
Bit Rate Mode | Variable (VBR) | Constant (CBR) | Variable (VBR) |
Target Bitrate | 2 - 6 Mbps | 4 - 10 Mbps | 1 - 4 Mbps |
FPS Rate | 25 - 30 | 30 | 15 - 30 |
Bandwidth | 4 Mbps | 6 Mbps | 4 Mbps |
Exposure Time/Shutter Speed | 1/60s - 1/250s | 1/1000s - 1/2000s | 1/30s - 1/120s |
Camera Specification
1. Crowd Estimation Extended References
In terms of Crowd Estimation, the resolution does affect accuracy. High resolution (e.g. HD 720p or Full-HD 1280p) is recommended for high density of crowd (>300 people in a frame) whilst smaller resolution (i.e.480p) is enough for low or moderate density of the crowd. Below is the table of recommendations for camera resolution, height, and installation angle.
| Camera Recommendations | High Density Crowd | Low to Moderate Density Crowd |
|---|---|---|
Resolution | 720p(1280x720) | min 480p (640x480) |
Height | 8-12m | 6-8m |
Angle | 30° - 40° | 30° - 40° |
Disclaimer
Model Capability
Nodeflux Crowd Estimation is focused on cases for a dense crowd. Below are some cases that are expected to be estimated correctly.
Currently, we also support low dense crowd similar to images below:
An improvement for Crowd Estimation model where we combine our People Dwelling and People Counting model to provide crowd estimation for density less than 100 people.
Result Interpretability
The Crowd Estimation model will produce two outputs which are heatmap points and estimation range.
The estimation crowd counts will be in min-max range (example: 120 - 300 people) instead of actual head counts.
Last updated
