AI Vision

AIVision is an AI-based video analytics platform that uses inputs from security cameras, USB cameras and video files.

It can identify over 26 classes of objects including fire, weapons, violence with the possibility of anonymizing blurred people and vehicles in live streams or videos.
The platform is based on an AI Framework and Proprietary Communication Protocol encrypted client server and P2P. Multiple processing devices can be added to the platform, starting from IoT devices to virtual processing servers, and on top of them the video streams to be processed can be added with different neural networks available in the platform. A Labels Management module is also implemented in the platform and is available in the customer account. It is connected with all the video streams added on the clients' processing devices.With this module you can create new object classes or improve the detection on existing classes. If clients already have an existing dataset they can use it for labelling privately without uploading the pictures to the AI Vision cloud.

The platform also offers new class training services that can be used to identify errors/defects in production workflows.

Using the Orange infrastructure, the processing and communication application is inserted over the edge-compute infrastructure linked to the private 5G network using URLLC slice (low latency, 8ms round-trip delay) and eMBB slice (very high throughput) alarms can be transmitted at very high speed (comparable or better than fiber) to devices connected to the private 5G network. Processing of traffic from customer rooms is done exclusively on-premise.