Verkada People Analytics User Guide
- August 8, 2024
- Verkada
Table of Contents
Verkada People Analytics
Product Information
Specifications
- Product Name: People Analytics
- Manufacturer: Verkada
- Features: Person of Interest notifications, Face search, Person attributes, Person history, Occupancy Trends
Product Usage Instructions
Responsible Use
People Analytics leverages computer vision and facial recognition
technologies. Users should be aware of the limitations and factors affecting
performance such as distance from the camera, image resolution, lighting, and
facial coverings.
Key Features
- Person of Interest notifications: Set alerts for specific individuals detected onsite.
- Face search: Search for individuals across all cameras based on facial matches.
- Person attributes: Search for people based on selected attributes like clothing color and gender appearance.
- Person history: View captured individuals to expedite investigations.
- Occupancy Trends: Measure foot traffic at critical locations.
Optimizing Analytics Performance
To ensure optimal performance, follow these guidelines:
- Head angle: Maintain a good face angle for accurate results.
- Distance to Camera: Ensure proper distance for person detection and facial recognition based on pixels per foot (PPF).
Calculating PPF Value
To calculate the PPF value at a specific distance: Divide the camera’s
horizontal pixel count by the width in feet of the camera’s horizontal field
of view (HFoV) at that distance.
FAQ
-
Q: How does Verkada ensure responsible use of People Analytics?
A: Verkada sets guardrails around the use of People Analytics, requiring opt- in and authorization by an organization’s admin for access. -
Q: What factors can impact the performance of People Analytics?
A: Factors such as distance from the camera, image resolution, lighting conditions, and facial coverings can affect the performance of People Analytics.
Overview
- Verkada’s People Analytics solution is designed to make it easier to find meaningful events and save time during critical investigations by enabling users to identify and search for individuals in your camera footage.
- Powered by Verkada’s industry-leading edge-based processing and computer vision technology, Verkada’s People Analytics solution includes face search, person attributes, person history, Person of Interest notifications, and Occupancy Trends.
- People Analytics delivers high-quality results with minimal bandwidth impact. All video, images, and data are encrypted at–rest and in–transit and stay within your Command Organization. To use Verkada’s People Analytics features, no additional software or hardware is needed. All Verkada camera models support People Analytics, except for the CD22 and CD22-E Dome cameras.
Responsible use
- People Analytics leverages computer vision and facial recognition technologies which are powerful tools that must be used responsibly. These technologies have limitations as they rely on the ability to interpret visual inputs. In addition, factors like distance from the camera, image resolution, camera installation and placement, lighting, and facial coverings like masks or hoods will affect performance.
- As has been widely reported, the accuracy of facial recognition technology may vary across diverse groups of people. At Verkada, we intentionally do not train our People Analytics models to identify sensitive characteristics such as race, ethnicity, age, or other demographic information.
- We also set administrative guardrails around the use of People Analytics: it is disabled by default and after opting in, only users authorized by an Organization’s admin may enable or view analytics on a given device.
Key features
What can you do with People Analytics? Once your camera is positioned correctly, explore all the use cases possible with this powerful toolkit.
-
Person of Interest notifications
Proactively set alerts for when a person is detected on site who has a face that matches a preselected individual. -
Face search
Search across all your organization’s cameras for people who have been detected as a match to an existing or uploaded face. -
Person attributes
Quickly search through your cameras to find people who match selected attributes, including clothing color, gender appearance, and more. -
Person history
View all people that have been captured by your Verkada cameras to speed up investigations. -
Occupancy Trends
Quantify foot traffic at critical locations.
Optimizing Analytics Performance
Ensure your cameras are positioned according to the following guidelines to get the most consistent results with People Analytics features.
Head angle
For optimal performance with face search and Person of Interest alerts,
cameras should capture faces head–on, facing the camera and as close to the
camera as possible.
The angle of captured faces must be within the following range of values:
Distance to Camera
- Person detection and facial recognition require a minimum level of detail, expressed quantitatively as pixels per foot (PPF). PPF decreases with distance from the camera. The PPF at any given distance varies with the camera’s field of view and the image resolution.
- To calculate the PPF value at the distance you wish to detect people, divide the camera’s horizontal pixel count (available in the camera’s datasheet under “Tech Specs” → “Sensor Resolution”) by the width in feet of the camera’s horizontal field of view (HFoV) at that distance.
This image was taken from a CD62 with a resolution of 3840 x 2160 and the width of the horizontal field of view at the target distance is 12 feet. Using the PPF formula we find that we have a PPF value of 320 at the doorway, which is sufficient for all People Analytics features.
The maximum target distances we recommend for our camera models are listed below. (These distances are recommendations only and can vary based on the conditions of a particular camera deployment.) Note that the PPF required to detect people and recognize faces depends on the camera model. Our latest generation camera models run on more powerful hardware and are therefore able to run larger, more powerful computer vision models. This allows them to detect people and recognize faces with fewer pixels per foot.
-
Person detection
(Person history, Person attributes) -
Facial recognition
(Face search, Person of Interest)
Camera Series
| Model Number| 0% zoom| 100% zoom| PPF| 0% zoom| 100% zoom|
PPF
---|---|---|---|---|---|---|---
Dome| | | | | | |
| CD22/E| 110.6ft / 33.7m| | 11| –| | –
| CD32/E| 110.6ft / 33.7m| | 11| 16.9ft / 5.1m| | 75
| CD42/E| 110.6ft / 33.7m| | 15| 22.1ft / 6.7m| | 75
| CD52/E| 100.0ft / 30.5m| 269.1ft / 82.0m| 15| 20.0ft / 6.1m| 53.8ft /
16.4m| 75
| CD62/E| 125.0ft / 38.1m| 409.3ft / 124.8m| 13| 30.0ft / 9.1m| 98.2ft /
29.9m| 54
| CD31/E| 11.3ft / 3.4m| | 75| 5.7ft / 1.7m| | 150
| CD41/E| 15.8ft / 4.8m| | 75| 7.9ft / 2.4m| | 150
| CD51/E| 19.9ft / 6.0m| 50.0ft / 15.3m| 75| 10.0ft / 3.0m| 26.8ft / 8.2m|
150
| CD61/E| 21.5ft / 6.5m| 75.0ft / 22.9m| 75| 10.7ft / 3.3m| 35.2ft / 10.7m|
150
| D40| 16.0ft / 4.9m| | 75| 8.0ft / 2.4m| | 150
Mini Dome| | | | | | |
| CM22| 108.7ft / 33.1m| | 11| | |
| CM42| 108.7ft / 33.1m| | 15| 21.7ft / 6.6m| | 75
| CM41/E/S| 22.1ft / 6.7m| | 75| 11.1 ft / 3.4m| | 150
| CM61| 31.6ft / 9.6m| | 75| 15.8ft / 4.8m| | 150
Bullet| | | | | | |
| CB52-E| 82.1ft / 25.0m| 253.0ft / 77.1m| 15| 16.42ft / 5.0m| 50.6ft /
15.4m| 75
| CB52-TE| 267.8ft / 81.6m| 637.5ft / 194.3m| 15| 53.6ft / 16.3m| 127.5ft /
38.9m| 75
| CB62-E| 111.3ft / 33.9m| 395.0ft / 120.4m| 13| 26.8ft / 8.2m| 95.1ft /
29.0m| 54
| CB62-TE| 395.0ft / 120.4m| 988.2ft / 301.2m| 13| 95.1ft / 29.0m| 237.9ft /
72.5m| 54
| CB51-E| 20.6ft / 6.3m| 50.0ft / 15.3m| 75| 10.30ft / 3.1m| 26.8ft / 8.2m|
150
| CB51-TE| 50.6ft / 15.4m| 125.0ft / 38.1m| 75| 25.3ft / 7.7m| 63.8ft /
19.4m| 150
| CB61-E| 23.5ft / 7.2m| 75.0ft / 22.9m| 75| 11.72ft / 3.6m| 33.3ft / 10.2m|
150
| CB61-TE| 66.7ft / 20.3m| 175.0ft / 53.3m| 75| 33.3ft / 10.2m| 85.6ft /
26.1m| 150
Fisheye| | | | | | |
| CF81-E (pano)| 40.0ft / 12.2m| | 40| 20.0ft / 6.1m| | 80
| CF81-E (4-way)| 18.0ft / 5.5m| | 75| 9ft / 2.74m| | 150
| CF81-E (ePTZ)| –| | 75| –| | 150
Multisensor| | | | | | |
| CH52-E| 103.1ft / 31.4m| 284.2ft / 86.6m| 15| 20.6ft / 6.3m| 56.8ft /
17.3m| 75
PTZ| | | | | | |
| CP52-E| 173.0ft / 52.7m| 4800.0ft / 1463.0m| 15| 35.0ft / 10.7m| 980.0ft /
298.7m| 75
Verkada’s CD22 and CD22-E do not support facial recognition.
Lighting
Lighting can impact People’s Analytics performance. Bright and even lighting will provide the most reliable and accurate results. Performance may suffer in low-light scenes as well as those with direct sunlight or high contrast. People Analytics generally performs best with a minimum light level of 250 lux, which can be measured with a standard lux meter.
Compliance and Availability
Verkada is committed to ensuring its hardware and features comply with applicable regulations. The following security camera features are not available in the jurisdictions listed below.
Jurisdiction
| Face detection| Gender appearance|
Person of Interest notifications
---|---|---|---
Illinois| X| X| X
Portland| X| X| X
Texas| X| X| X