‍

Challenge

‍

Docler Holding is a multinational enterprise, headquartered in Luxembourg since 2013, which counts more than 1000 employees worldwide. The company started in 2001 as a garage project, powered by the ideas and visions of young and enthusiastic Hungarian entrepreneurs. Today, Docler Holding has created and developed a large number of highly diversified companies, all gathered under a unique umbrella. These businesses perform and experience ongoing growth in the fields of technology, media, entertainment, sports and art.

Docler Holding is the operator of one of the most visited websites in the world, based on cloud-based video streaming. Content curation and moderation is a resource intensive activity involving human operators. However, recent advances in machine learning and computer vision have created an opportunity to automate and enhance such tasks, which has significant business potential.

‍

Solution

‍

To address this challenge, the Research and Development Department at Docler partnered with IncQuery Labs, in order to create mSys, a cloud-based machine learning computer vision system for automated content classification, quality analysis and moderation for real-time video streaming. The main goal of mSys is to provide a modular, extendable cloud-based architecture for high-performance video stream analysis, leveraging state-of-the-art computer vision and AI algorithms, for a variety of analysis tasks based on object tracking, facial recognition, pose tracking, object classification, and others.

‍

Result

‍

Working together for more than 18 months in close collaboration, Docler R&D and IncQuery Labs have prototyped a complex cloud-based solution. Taking full advantage of Microsoft Azure and its ML capabilities, mSys provides support for both developing and training custom AI models, as well as running them in a production environment that can scale to the requirements of the live website.  

As a common feature to both the development and live environments, mSys provides custom-built database solutions necessary for developing custom AI models, and also includes a CI/CD pipeline based on Jenkins and Docker. The live environment is based on a bespoke engine built using C++ for high-performance processing.

In the final phase of the collaboration, the experts of IncQuery Labs have assisted Docler R&D to further develop mSys into a full-featured internal product, by transferring knowledge and providing takeover support for both software engineers as well as devops specialists.