DX Play Computer Vision

Powerful AI solutions for WebRTC stream processing

One of the key features of the DX Play product is the ability to ingest streams from multiple sources, process them with CV/ML models made for specific purposes, and display the streams with AI image processing along with all the necessary metadata in a convenient user interface. This represents one of the most efficient and powerful AI solutions on the market. Let’s dive deeper into how the system works and what benefits of AI for business it can bring.

AI solutions for real time streaming and processing

Stream Sources

The system is able to ingest multiple stream formats from different sources. Those can be CCTV cameras, user webcams, GoPro streams from an RTMP server, WebRTC streams, or any other kind of video stream that is being supported by Gstreamer. Speaking about Gstreamer, it’s capable of receiving a lot of stream formats, basically, all the popular ones available at the moment. The horizontal AI scalability of the DeepXHub platform enables multiple simultaneous streams.

Real Time AI Image Processing

The first module of the system responsible for receiving user streams is Gstreamer. It’s fast, reliable, highly flexible, and modular allowing for multiple format handling and the most effective resource usage available at the moment.
Stream handling modules are tightly connected with computer vision and machine learning models, which are used for processing streams and generating high-level metadata. The integration of CV/ML models into the media pipeline provides the ability to do real time processing with AI. The main benefit of using the system is the ability to monitor a large number of streams 24/7, with constantly improving precision as the models are being enhanced over time. This is where AI solutions can outperform human workers in terms of cost-effectiveness and accuracy, covering a wide range of use cases for business process automation.

Metadata streaming in AI solutions

During their performance computer vision and machine learning models output metadata. This is high-level user-friendly data, which can contain multiple object detection, important events, critical parameter values, etc. In order to transfer this data to the user interface our system uses a set of high-performance data transmission tools.
The main processing module has Kafka producer, which is connected to the message cluster. Kafka consumers connect to message cluster and get the real-time metadata, which is further transmitted to the user interface with Websockets. The usage of Websockets gives us high performance and lowest latency possible, which is crucial for real time monitoring.

WebRTC Video Streaming

The processed stream with all the necessary annotations and highlights is streamed to the user interface with the help of WebRTC technology. WebRTC video streaming is realized with a separate Gstreamer pipeline, requiring communication with signaling server for setting the initial connection with front-end by P2P protocol. WebRTC live streaming enables high performance and low latency, while Gstreamer provides great capabilities for tuning the stream quality and vertical scalability.


The scalability of our system is realized with the help of Kubernetes (streaming and AI processing part) and Kafka (metadata transferring part).
From the user’s point of view, it results in a seamless experience when new streams are added to AI image processing. If the power of one instance on the back-end reaches its limit, a new one is automatically allocated for processing newly added streams.
The growing number of streams and processing modules inevitably results in larger amounts of metadata being generated. The scalability on this side is covered by Kafka architecture, which is also highly scalable.

Computer Vision platform benefits

To sum everything up, let’s go through the list of benefits our system can provide for business process automation, improvement and optimization:

  • Multiple stream formats supported
  • High performance, reliable and cost effective AI processing
  • Low latency of video and metadata transferring
  • Custom computer vision and machine learning models for any business needs
  • High scalability

About DeepX

DeepX is an R&D-intensive and innovation-driven consortium led by DEEPXTECH S.R.L. Italy, Covijn Ltd (“Co-vision project”) the United Kingdom, and AIMech Technologies (“Omnidetector” product) the United States, that provides Artificial Intelligence-powered Computer Vision solutions for businesses, empowering medium-sized and large enterprise companies with advanced AI-driven solutions since 2018.

At DeepX, we synergize our multi-year expertise to dive deep into each business’ specifics and achieve measurable business benefits. With our continuous research and a history of successful computer vision software development projects, we offer highly specialized expertise in some of the most challenging fields of computer vision.

Our services include, but are not limited to the following:

  • Computer Vision custom web & software development
  • Computer Vision products integration
  • Video intelligence-based applications development
  • AI-Powered real-time communication implementation and development
  • Machine Learning and Artificial Intelligence systems design & development
  • Computer Vision and Video Intelligence consulting
  • Artificial Intelligence and Machine Learning consulting

If you’re looking for new opportunities, put your trust in our vast experience and open the door to new possibilities!

Our expert team will guide you through the entire process of implementing technological solutions in accordance with all standards. Contact our experts for a free consultation on the potential use of AI in your business.

Close Bitnami banner