patent: split ai

Optimize available bandwidth for efficient video uploads

unnamed

Split AI is a patented technology that enables Dragonfruit to provide WAN optimization for high-resolution video uploads. While there is already a considerable AI engine in the Dragonfruit cloud, with Split AI there are also AI capabilities on-premise in the Dragonfruit Client. Using the processing power on the customer server, the Dragonfruit Client performs AI inference on-prem, detecting key objects of interest as well as ancillary objects. The core objects are transferred to the cloud at high-resolution, while secondary objects are downsampled before being uploaded, thereby conserving bandwidth. In the cloud, the key objects can be searched, counted, tracked, investigated, etc., as you would any other video object.

Why Dragonfruit patented Split AI

Bandwidth is a challenge

Bandwidth is one of the greatest challenges for uploading video to the cloud

WAN optimization

WAN optimization is critical to leverage the power of the cloud

Cloud x On-prem

Dragonfruit’s Split AI technology runs AI in the cloud as well as on-premise

How Split AI Works Example

With Split AI
Without Split AI

Without Split AI

For example, in a typical situation a CCTV camera may face a parking lot with lots of objects — cars, street lights, trees, mountains, etc.

With Split AI

Split AI allows the critical objects — in this case cars — to be uploaded at full resolution, while downsampling the non-critical objects in the background such as trees and mountains

Without Split AI

For example, in a typical situation a CCTV camera may face a parking lot with lots of objects — cars, street lights, trees, mountains, etc.

With Split AI

Split AI allows the critical objects — in this case cars — to be uploaded at full resolution, while downsampling the non-critical objects in the background such as trees and mountains

pasted image 0

High-resolution objects

In the cloud, you can run full analytics on the uploaded high-resolution cars such as extracting information such as make/model, vehicle color, license plates, etc. Split AI allows the processing to be split between the cloud and client to optimize bandwidth while still yielding the most accurate analytics results.