Frequently Asked Questions (FAQs)

1. Are we a software or hardware company?

We are a software company. We do not sell cameras or edge devices but have partners who can assist you in procuring hardware.

2. What kinds of cameras can we work with?

To learn more about supported video streams please visit Read Me

3. Do we help with installation?

We do not assist with the installation of cameras or edge devices, but we do offer technical support for setup and use via email, Slack, and Zoom.

4. Prerequisites to run Lanthorn?

Lanthorn prerequisites
See footnote for a list of linux operating systems.1

5. What is the high-level architecture of Lanthorn?

The architecture of the Lanthorn can be split into 2 components: Lanthorn processor & dashboard.

Lanthorn Processor - Lanthorn can run on Linux OS on the cloud or edge devices (on-premise Graphic Processing Unit - GPU). The processor (GPU) is the unit responsible for video processing. See processor performance for detailed information.

Lanthorn Dashboard - A web application that helps you to set up your cameras and processor. The dashboard also visualizes and analyzes the results generated by the processor.

6. What is Edge vs Cloud Computing?

Edge computing takes place on an edge device, which is a piece of hardware typically located on premise. Edge computing is used to process time-sensitive data, while cloud computing is used to process data that is not time-driven. Edge computing also keeps all data on premise and in your hands, while cloud computing can risk exposing your data.

7. What data is generated by Lanthorn?

Lanthorn’s computer vision engine can detect people and count them to provide data on occupancy. It can also detect objects including fask masks and conduct pose estimation to detect details such as social distancing. For more information on how the data generated by Lanthorn can be used, visit Use Cases

8. How does Lanthorn preserve data security?

None of your data will leave your servers unless you choose to export it. Since the processor can be deployed on lightweight edge devices,you do not need to expose any of your camera streams outside of your local network (LAN); you only need to include the device inside the network. This configuration guarantees that the processing of the information stays inside of your private network; without transmitting anything to the cloud. For more information, please see: Privacy & Security whitepaper

9. How does the API work?

The way that a dashboard (and any other third-party application) interacts with the processor is through API endpoints (documented here). You don’t need to expose these endpoints outside your network, as the browser interacts directly with the processor. If you try to log in into the dashboard outside the network, you will receive the following error:

Lanthorn prerequisites

10. How does Lanthorn protect occupancy privacy (GDPR)?

We are aware that the processed videos probably contain sensitive information; such as the identity of the people recorded. Taking into account that the input of Lanthorn is videos, maintaining them inside the network is important to keep the privacy of the people recorded. The process doesn’t require face recognition to accurately detect occupancy. In addition, all the data generated and stored is also anonymized to respect personal privacy; storing only statistical information, such as the number of people.

11. Does Lanthorn use facial recognition?

We do not use facial recognition. The model detects that a person is standing because it detects a “person-like object” (not necessarily a silhouette).

12. Is communication Encrypted?

All communications between Lanthorn’s dashboard and processor(s) are encrypted.

13. What parts of Lanthorn are open source?

Lanthorn’s occupancy computer vision engine is open source; however, our API engine and dashboard are not.

14. What parts of Lanthor are customizable?

The Lanthorn dashboard is customizable and can be white-labeled for customers who prefer to have their own branding. New metrics and visualizations can also be added depending on the customer’s needs and requirements. Moreover, new computer vision models (other types of detections) can also be added to our application. Custom computer Vision model development needs data and time for training and accuracy improvement.

15. What is the model accuracy and performance on different hardware?2

This sheet will serve as a guide on how different devices and models perform.

16. How much does Lanthorn cost?

We charge per camera stream. Please visit for specific pricing.

17. Where can I find technical information and guidance about the installation?

Please visit Read Me


1 NVIDIA Jetson Nano; NVIDIA Jetson TX2; Coral Dev Board; AMD64 node with attached Coral USB Accelerator; X86 node; X86 node accelerated with OpenVino toolkit; X86 node with Nvidia GPU

2 Processor performance