Videoloft’s people counting model

We count the people entering and exiting your site, so you get straightforward insights that support better, more confident decision-making.

How it works

Multiple frames showing a person entering a retail store through glass doors, illustrating movement detection across video frames.

Multiple still frames are extracted from the video and analyzed to detect people.

Person entering a store highlighted with a yellow detection box and 100 percent confidence indicator.
The detected people are marked with bounding boxes and a confidence score.
Counting line marked at the store entrance threshold showing where entries and exits are measured.
Results are then checked against a user-defined count line. Counts and direction are recorded when people cross the line.
Videoloft people counting dashboard displaying people counting results
Results are displayed in the people counting dashboard for easy analysis.

How its trained

Our current people counting model is based on a standard YOLO framework and has not yet been specifically trained on Videoloft-specific footage. Soon, we’ll begin custom training using real-world footage from cameras opted into the Videoloft analytics program, which will improve accuracy further.

Model releases

Each release is tested extensively and benchmarked. Below you’ll find version history, accuracy scores, and known limitations.

V1 Initial release Oct 24 92% accuracy

Release Date October 2024

Summary Production release V1 of people counting Model.

Accuracy results

Metric Accuracy
People counted entering location
93%
People counted exiting location
91%
Overall accuracy
92%
We test each version of our model against a set of real customer video. Let’s explain what these numbers mean…
 
  • Release V1 detected and counted 93% of people who entered locations.
  • Release V1 detected and counted 91% of people who exited locations.
  • Therefore the overall detection and counting accuracy is 92%.
There are some situations where the model has limitations. The example images below are AI-generated to protect customer data.

Known limitations

Group of shoppers entering a retail store together, partially blocking one another from the overhead camera view.
Individuals being partially hidden by other members of a group and not being detected.
Indoor security camera view of a reception area with a coat hanging on a stand near the entrance, highlighted to demonstrate how people counting systems can sometimes create false positives by detecting objects as people.
False positive detections of objects that are not people (e.g coat hanger by a door).
Retail store entrance at night with glare from exterior lighting and reflections on glass doors, reducing visibility for people detection.
Missed detections or false detections particularly in low light or night conditions.

hello@videoloft.net

Toll-free US: (855) 737-0237
UK: +44 203 9870133

+44 753 280 1187

Try it for free

For more information, contact us directly or fill in your details using our quick form and a representative of our global sales team will be in touch.

Your details

We value your privacy and take data protection seriously. We will never share or sell your personal data to third parties. Your trust is important to us, and we’re committed to safeguarding your information.