AI – Smart Social Distancing

Social distancing is one of the most important defenses against the spread of COVID-19. We bring here a open source based solution for ”Smart Social Distancing with AI” application. This open-source application based on Jetson Nano helps businesses monitor social distancing practices on their premises and take corrective action in real time. 

The application’s main tasks are done by the “Computer Vision Engine” module. This module manages the pre-processing, inference, post-processing and distance calculations on the data. The detection is done by a SSD-Mobilenet-v2 convolutional neural network running on Jetson Nano and optimized using the NVIDIA TensorRT accelerator library. 

A dashboard, generated by the GUI and web application module, displays results that include social distancing violation alerts by location, color-coded bounding boxes to convey the extent of the violation, and overall violation trends. These alerts and visuals help a business take appropriate action. 

The application can be customized or extended using parameters provided in the configuration file or source code. For example, developers can use their own network models or add a custom video file or stream as inputs to quickly get started building and deploying their own Smart Social Distancing applications.

The roadmap includes several enhancements, adding face mask detection, adding safety score for a site, evaluating other network models and adding more robust distance calculation methods.

We at Unizen Technologies with our expertise on AI-based technology provide end to end solution on solving complex problems with cutting edge technologies.

This source is based on open source community in the GitHub repo

AI at Edge

AI at the Edge : Saving Bandwidth with Anomaly Detection

Most smart city applications today rely on analyzing large amounts of video data from cameras.

The ability to identify and reason over the most relevant events within a video is essential to build efficient and scalable applications.

The NVIDIA Jetson Nano-based brings in the Intelligent Video Analytics and Smart Cities category all offline at edge

At the heart of the application is an auto-encoder model trained and running on a Jetson Nano using TensorFlow and Keras. This model learns the context of a scene with each oncoming frame of the video and develops the ability to flag anomalous events. The team proposes that these anomalous events can be processed using DeepStream SDK for additional inference–for example, to identify and track objects in the scene.

In a scene with continuous activity such as a busy road, this method of tracking the structure of the image and flagging major changes to the image as anomalies is better than a simple motion detection algorithm. In the video below, the application correctly flags anomalous events and in the process reduces the image feed by 100X.

Model correctly identifies an anomaly event that can be further analyzed

This model can be used within a video analytics pipeline to build smart city applications that make optimal use of network and cloud resources

At Unizen Technologies, we have expertise to build AI model and customize and train the data for Production Solutions.

We have expertise in Edge Analytics based on Nvidia Jetson platform