Frequently Asked Questions

Our video analytics software is compatible with any type of video camera, regardless of the brand or model. Camera quality is not a major concern for our software, as we have trained computer vision models that are able to detect people and objects even under conditions of poor image quality or low lighting. While integration of IP cameras is straightforward, we can also support analog cameras that are connected to a DVR, as well as USB cameras that are connected to a local computer or embedded boards.

We take privacy and data security very seriously and implement our solutions using best practices. For customers with strong security and compliance requirements, we offer on-site solutions that can operate on servers provisioned by their own company IT departments. Additionally, we can operate our solution so that no video is recorded on the server storage. In cases where videos need to be retained as proof in case of an accident or incident, for example, we can redact the videos to remove personal identifiable information (PII) by blurring faces and license plates.

Stura specializes in computer vision solutions that detect people in a variety of scenarios, such as retail spaces, office buildings, and open spaces. As a computer vision company, we have the ability to train models that are capable of detecting categories of people based on their general appearance. For example, in people-counting applications for stores, we can train our model to differentiate between regular visitors and staff (assuming staff members are wearing uniforms). In general, our models are able to detect any type of objects (vehicles, bicycles, pieces of baggage, etc). We can also integrate technology for facial recognition or license plate recognition.

We offer a variety of options for data storage. Many customers are happy to access their data via CSV files. We can also store your data in a database table of your choice. See this answer for additional details.

We design our solutions to meet our customers’ requirements and make integration seamless. Most customers prefer to receive their analytics data in the form of simple CSV or XLSX (MS Excel) files. Other customers integrate our data via calls to REST API endpoints or IPC (inter-process communication), for example, by using REDIS streams and pub/sub mechanisms or other RPC solutions.

Our software can be configured for real-time analytics. It can trigger notifications when certain conditions are met, such as when the wait time at a cashier surpasses a predetermined threshold. In some situations, real-time data or notifications are not needed, and therefore, we can configure our software for post-event analysis (also called ‘batch processing’). Batch processing of the videos allows for optimization of computing resources. For example, in a retail setting, while stores are normally open for 10 to 12 hours per day, we can use an on-premise server for the entire 24 hours, thereby processing twice the number of cameras that would normally be supported if the data were to be processed in real-time.

During the setup phase, we validate the accuracy of our model by comparing their output to data that can be extracted manually on video samples that are representative of the operating conditions. We periodically review the accuracy of our software to make sure there is no degradation in accuracy in case of changes in the environment in which the software operates. Additionally, our software will export annotated videos that can be inspected by our customers to verify that the algorithm is functioning properly.

Stura handles projects of various sizes, including computer vision with one camera or deployments with thousands of cameras. While there is no limit to the number of cameras, bandwidth and server allocation must be considered. For example, when processing videos on the cloud, sufficient bandwidth is needed to stream videos, and data usage costs might be an issue (e.g., cellular connections). On-premise deployments don’t have bandwidth limitations, but several servers may be needed to process all the cameras. Both bandwidth usage and computing power can be optimized via a carefully designed solution. We are committed to working with our customers to find the ideal balance of accuracy and resources for successful projects with minimal long-term cost.