Open Source Smart Camera to Detect Face Masks, Build MaskCam with Tryolabs and Seeed’s Edge Devices for NVIDIA Jetson

Challenge

During the pandemic, minding if people were using masks indoors or outdoors started to be the typical scenario. And even though making this identification by a human eye seems to be a trivial task, it ended up being an interesting challenge for the current development of Artificial Intelligence systems. Unlike facial recognition systems, this solution would not need to identify the exact person with features such as eye, nose, mouth, ears. Still, it can determine the correct use of a mask in whatever scenario or facial characteristic is present. 

“If we can compute the number [of people who are complying with the mask mandates], people can make policies and monitor on whether or not they need to do another campaign to push mask usage,” says Alan Descoins, the Chief Technology Officer of Tryolabs.

In March 2021, MaskCam was demonstrated and introduced by the joint efforts and leading technology from BDTI, Jabil, NVIDIA, and AI and Machine learning solutions company Tryolabs develop AI and Machine learning solutions capable of detecting people passing through a security-like camera and identifying if they were using a face mask or not. Now we are also glad to apply this great open-source project MaskCaM with Seeed’s Jetson Platform edge devices.

As part of a multi-partner team with engineering consulting firm BDTI, AI solution and GPU provider NVIDIA, and electronics manufacturer Jabil, this AI and Machine learning solutions company, Tryolabs developed a system capable of detecting people passing through a security-like camera and identifying if they were using a face mask or not. Now we are also glad to apply this great open-source project MaskCam with Seeed’s Jetson Platform edge devices.

MaskCam: an Open Source Smart Camera to Detect Face Masks

The Face Mask Solution powered by Tryolabs is a prototype reference design for an NVIDIA Jetson module-based smart camera system that measures crowd face mask usage in real-time, with all AI computation performed at the edge. We have deployed MashCam our A206 Jetson Carrier board. Combining the USB camera, the smart camera detects and tracks people in its field of view and determines whether they are wearing a mask via an object detection, tracking, and voting algorithm. It uploads statistics to the cloud, where a web GUI can be used to monitor face mask compliance in the field of view. Please check the below demo video.

Tryolabs

Tryolabs logo

Tryolabs is with the mission of being the leading platform for AI transformation. Tryolabs team creates high-impact solutions by applying computer vision, natural language processing, predictive analytics, and more techniques in the AI field. We are partnering with Tryolabs to solve complex problems together, tailored to individual business cases, leveraging Seeed’s decade’s hardware expertise and Tryolabs’s unique AI solution. On their website, you can find solutions for:

Face mask detection in street camera video streams powered by Tryolabs and Seeed Hardware for NVIDIA Jetson

The on-device software stack is mostly written in Python and able to run under JetPack 4.6. Edge AI processing is handled by NVIDIA’s DeepStream video analytics framework, YOLOv4-tiny, and Tryolabs’ Norfair tracker. Tryolabs analytics offers report statistics to and receives commands from the cloud using MQTT and a web-based GUI. The software is containerized and for evaluation can be easily installed on an NVIDIA Jetson SoMs using docker with just a couple of commands. The device can also run under remote management such as mender.io, balenaOS, which makes it easy to manage and deploy multiple devices.

Tryolabs face-mask recognition technology overlaid on EarthCam CCTV of Temple Bar in Dublin, Ireland
Tryolabs’ software identifies whether visible faces are uncovered or wearing masks as they pass by a closed-circuit television camera near Temple Bar in Dublin, Ireland. CCTV footage collected by EarthCam; analysis courtesy of Tryolabs.
Tryolabs face-mask recognition technology overlaid on EarthCam CCTV of Little Italy, New York City
Examining CCTV footage with mask recognition could offer a sense of how many people are compliant with health guidance in a particular area, such as the Little Italy neighborhood of New York City. CCTV footage collected by EarthCam; analysis courtesy of Tryolabs.

Hardware components

Finding a piece of hardware that can satisfy the requirements of the AI application is not easy. It required a compact size and cost-effective computer that can run object detection models in real-time.

NVIDIA Jetson Powered Solution

Face Mask solution is built around the NVIDIA Jetson  SOM (system-on-module), which somehow manages to fit a CPU, GPU, RAM, and flash memory into a bit smaller than a credit card. Here we will recommend the Seeed A206 Jetson Carrier board, with a camera and a must-have heatsink to use with Jetson Nano SoM.

You can also choose Jetson SUB Mini PC-Blue with Jetson Xavier NX module, an Aluminium case with cooling fan, 128GB SSD, WiFi, Antennas, and pre-installed JetPack System.

Mask Cam supports the following Jetson platform hardware

Seeed Nvidia Jetson SoMs powered Edge Devices

Seeed Nvidia Jetson SoMs powered Edge Devices

Seeed A206 Jetson Carrier board

Seeed’s newest A206 carrier board has nearly the same design and function as NVIDIA® Jetson Xavier™ NX carrier board, fully compatible with Jetson Nano/Xavier NX/TX2 NX module. It provides several connectors with industry-standard pinouts to support additional functionality beyond what is integrated on the main platform board.

The board provides HDMI 2.0, Gigabit Ethernet, USB 3.0, M.2 key E wifi / BT, M.2 key M, CSI camera, RS232, CAN, PIO, I2C, I2S fans, and other rich peripheral interfaces, which are compatible with multiple NVIDIA SoMs and offer a wide variety of options and flexibility.

Seeed’s Nvidia Jetson Compatible Carrier Boards

Camera module

Raspberry Pi High-Quality Camera can be a good candidate, which is built around the IMX477 image sensor, which supports 4032×3040 resolution at 30 FPS and 1920×1080 resolution at 60 FPS. Compare with Raspberry Pi Camera Module V2 also based on Sony IMX219, new cameras for Jetson Nano and Jetson Xavier NX starts at a cheaper price at $18.90. The 8MP Infrared Night Vision Camera with 77° FOV with double IR LED modules only starts from $20.90. IMX219 cameras for Jetson Nano and Jetson Xavier NX also include different degrees of Field of View (FOV) functions to help the camera image capturing avoid edge vignettinPlease refer to Seeed blog on How To Choose Cameras For NVIDIA Jetson Nano and Xavier NX.

Nvidia Jetson Compatible Accessories

Join Seeed Partner Ecosystem

We are looking for partners to join our ecosystem together to deliver solutions to different industries together. By consolidating the latest and the best technology, Seeed aims at emerging all kinds of enterprise scenarios. Combining Seeed’s best-in-class hardware, the software platform from our partner ecosystem, and community knowledge, developers and system integrators will find the easiest ever way to consolidate the IoT solution to their existing or new systems. 

Please contact [email protected] for more partnership ecosystem information.

Enterprises strive to deliver more and the best product value to customers. The unique and robust feature provided by the Internet of Things ecosystem focus on the maintained and continuous close relationship between IoT platform and their terminal users. We will be glad to partner with cloud services, software platforms, and superior ISVs for:

  • Delivering ready-to-deploy and co-integrated solutions to global companies and system integrators. 
  • Delivering added value for verticals audience
  • Integrating partners unique solution into Seeed hardware platform and sharing with community together.
  • Lowering the threshold for end users to from quickly POCs to scale industrial scenarios.

1 thought on “Open Source Smart Camera to Detect Face Masks, Build MaskCam with Tryolabs and Seeed’s Edge Devices for NVIDIA Jetson

Comments are closed.

Calendar

January 2022
M T W T F S S
 12
3456789
10111213141516
17181920212223
24252627282930
31