Transparency note for Spatial Analysis
This Transparency Note discusses Spatial Analysis and the key considerations for making use of this technology responsibly.
What is a Transparency note?
An AI system includes not only the technology, but also the people who will use it, the people who will be affected by it, and the environment in which it is deployed. Creating a system that is fit for its intended purpose requires an understanding of how the technology works, its capabilities and limitations, and how to achieve the best performance. Microsoft's Transparency Notes are intended to help you understand how our AI technology works, the choices system owners can make that influence system performance and behavior, and the importance of thinking about the whole system, including the technology, the people, and the environment. You can use Transparency Notes when developing or deploying your own system, or share them with the people who will use or be affected by your system.
Microsoft's Transparency notes are part of a broader effort at Microsoft to put our AI principles into practice. To find out more, see Responsible AI principles from Microsoft.
This Transparency Note discusses Spatial Analysis and the key considerations for making use of this technology responsibly. There are a number of things you need to consider when deciding how to use and implement AI-powered products and features:
- Will this product or feature perform well in my scenario? Before deploying AI into your scenario, test how it performs using real-life data and make sure it can deliver the accuracy you need.
- Are we equipped to identify and respond to errors? AI-powered products and features will not be 100% accurate, so consider how you will identify and respond to any errors that may occur.
Introduction to Spatial Analysis
Computer Vision spatial analysis is a new feature of Azure Cognitive Services Computer Vision that helps organizations maximize the value of their physical spaces by understanding people's movements and presence within a given area. It allows you to ingest video from CCTV or surveillance cameras, run AI operations to extract insights from the video streams, and generate events to be used by other systems. With input from a camera stream, an AI operation can do things like count the number of people entering a space or measure compliance with face mask and social distancing guidelines.
For more information, see Overview of Spatial Analysis.
Example use cases for spatial analysis
The following are example use cases that we had in mind as we designed and tested spatial analysis.
Social Distancing Compliance - An office space has several cameras that use spatial analysis to monitor social distancing compliance by measuring the distance between people. The facilities manager can use heatmaps showing aggregate statistics of social distancing compliance over time to adjust the workspace and make social distancing easier.
Shopper Analysis - A grocery store uses cameras pointed at product displays to measure the impact of merchandising changes on store traffic. The system allows the store manager to identify which new products drive the most change to engagement.
Queue Management - Cameras pointed at checkout queues provide alerts to managers when wait time gets too long, allowing them to open more lines. Historical data on queue abandonment gives insights into consumer behavior.
Face Mask Compliance – Retail stores can use cameras pointing at the store fronts to check if customers walking into the store are wearing face masks to maintain safety compliance and analyze aggregate statistics to gain insights on mask usage trends.
Building Occupancy & Analysis - An office building uses cameras focused on entrances to key spaces to measure footfall and how people use the workplace. Insights allow the building manager to adjust service and layout to better serve occupants.
Minimum Staff Detection - In a data center, cameras monitor activity around servers. When employees are physically fixing sensitive equipment two people are always required to be present during the repair for security reasons. Cameras are used to verify that this guideline is followed.
Workplace Optimization - In a fast casual restaurant, cameras in the kitchen are used to generate aggregate information about employee workflow. This is used by managers to improve processes and training for the team.
Considerations when choosing a use case
Avoid real-time critical safety alerting - Spatial analysis was not designed for critical safety real-time alerting. It should not be relied on for scenarios when real-time alerts are needed to trigger intervention to prevent injury, like turning off a piece of heavy machinery when a person is present. It can be used for risk reduction using statistics and intervention to reduce risky behavior, like people entering a restricted/forbidden area.
Avoid use for employment-related decisions - Spatial analysis provides probabilistic metrics regarding the location and movement of people within a space. While this data may be useful for aggregate process improvement, the data is not a good indicator of individual worker performance and should not be used for making employment-related decisions.
Avoid use for health care-related decisions - Spatial analysis provides probabilistic and partial data related to people's movements. The data is not suitable for making health-related decisions.
Carefully consider use in public spaces - Evaluate camera locations and positions, adjusting angles and region of interests to minimize collection from public spaces. Lighting and weather in public spaces such as streets and parks will significantly impact the performance of the spatial analysis system, and it is extremely difficult to provide effective disclosure in public spaces.
Avoid use in protected spaces - Protect individuals' privacy by evaluating camera locations and positions, adjusting angles and region of interests so they do not overlook protected areas, such as restrooms.
Carefully consider use in schools or elderly care facilities - Spatial analysis has not been heavily tested with data containing minors under the age of 18 or adults over age 65. We would recommend that customers thoroughly evaluate error rates for their scenario in environments where these ages predominate.