Computer Vision: Gemmo Video Platform

Gemmo Video Platform Flexible and scalable, the latest innovation from Gemmo is the Gemmo Video Platform. The SaaS Video Analysis platform, powered by the newest breakthroughs in Computer Vision. It’s an open archi...

Computer Vision: Gemmo Video Platform

Gemmo Video Platform

Flexible and scalable, the latest innovation from Gemmo is the Gemmo Video Platform. The SaaS Video Analysis platform, powered by the newest breakthroughs in Computer Vision. It’s an open architecture cloud based system that manages different video analytics algorithms on a single platform. It extracts profitable insights from live and recorded video clips. In other words, it automatically produces meaningful and relevant information from digital video that would have required human monitoring just a few years ago.

How does it work

All you have to do is upload a video on the platform or connect it to a live camera. The platform analyzes any video and gives you insights. The basic tasks performed are:

  • Image Classification: select the category of an image among a set of predetermined categories (e.g. car, person, ball).
  • Event Detection: classify specific event in the video sequence
  • Localization: locate an object in an image
  • Object:
    • Detection: categorize an object in an image.
    • Identification: given a target object, identify all of its instances in an image
    • Tracking: track the movement of an object within the video frame.


The Intelligent video analytics software has numerous applications in both the public and private sectors. Despite the variety of uses, the common denominator is that a vision analytics system permits human operators to bypass tedious rote tasks. It allows you to extract information from a great volume of stream video, with little effort. It also proves to be more accurate, reliable, and 100% functioning.

So, let’s see all sectors of applications for which the platform was designed.

Safety & Security

The Sparkd Video Platform leverages existing video surveillance networks to derive intelligence insights from live or recorded video content. It detects and extracts objects and human presence in video streams. It classifies them by their features and their interaction with the environment. Therefore it is possible to configure the system to trigger alerts for specific events and incidents before they happen or post-episode to facilitate investigation.
The possible domains that it covers are vary and applicable both for home and public places security.


  • Recognition, tracking and tracing of Person of Interest and event alert (Defense and Intelligence purposes)
  • Detection of suspicious behavior or identifying injurious objects to prevent crime
  • Body and face attributes detection and indexing (useful for example in institutional centers or organizations)
  • People counting in public spaces (airports, convention centers, trade fairs,…)
  • Human traffic flow
  • Crowd dispersion and direction detection (public events, concerts, conferences centre,…)
  • Queue detection (shopping centre, airports, supermarket checkouts,…)
  • Vehicle recognition
  • Monitor traffic jams


Sport analytics 

As widely discussed in AI in Sport Tech – The Lumascape, automated video analysis has an increasing popularity in the sport industry sector. More and more clubs and federations, coaches and athletes across all sports are taking advantage of computer vision. It allows them to obtain accurate performance metrics of the players and get a full picture end-to-end understanding of a match (player positions, ball passes, scores etc.) both in real time or uploading a recorded video clip. It is a valuable instrument to support coaches’ decision making process (tactics, strategies, pinpoint failures, tackle opponents game style). And it can help scouting agencies to evaluate players. But it is not a technology that applies only to professionals, even amateurs can take a critical look at their own performance in order to improve skills and prevent injuries.

These sport analytics platforms are also game changer for the automatic generation of sports video highlights.
Let’s now take a look deeper into the three main features that the Sparkd Video Platform offers for sport videos.


  • Event tagging / automatic highlights
    The Sparkd Video Platform integrates a cloud service for video storage with a video management solution which includes a set of manipulation and analysis techniques. The Sparkd Automatic Video Tagging is capable of assigning multiple tags to resources for listing and managing your media library. Leveraged with Machine Learning approaches that permit the visual recognition of objects and people, the platform can automatically identify scenes/events and tag them in a few seconds.
    You can create clips for any action of interest which can be watched back at any time by simply clicking on the register of that action in the timeline. You can even tag a match whilst watching it live.
  • Virtual Camera 

Auto Zoom and tracking is a feature that enables your camera to digitally zoom in on a moving object to capture details and track its movement autonomously. It is vital to ensure that important moments and movements in each game are adequately recorded.

  • Video Analytics

From the analysis of the videos it is also possible to derive statistical insights and information and consequently generate reports.


Similar to sport analysis, data is essential to understanding how to optimize virtual game play. Like in sports, gamers need to understand the best strategies and techniques to prevail against rivals, and consequently, ensure victory. The increasingly common trend to save the video of long-hours-Twitch-streaming allows users to re-watch the game and examine the performance further. If uploaded on video analytics platforms, the data gathering process automatically provides a dashboard with scoring and graphics about past matches.
One of the factors that contribute to the growth of E-sports in the last years is the consistent growth in viewers for E-sports competitions, which consequently attracted investments, advertisements and a great media coverage. Various gaming communities demand for immediate and automatic highlights of best gaming moments. That’s where these platforms come in handy.


  • Video summation and highlighting of video streaming into clips (key moments, clutch kills, match-winning plays)
  • Analysis and identification of successful patterns to improve the game play and minimize fails (e.g. main areas of the pitch, ball possession, tactics)
  • Storage and management of game play recordings sharable at any moment with your community



Manufacturers have started relying completely on digital technologies. This means also the adoption of cameras integrated with video analytics software to constantly monitor the conveyor belt and detect how each product is produced and assembled. Video analytics platforms help prevent failures through a warning system that reports in real time the identification of anomalies or damaged products. It is then possible to set them aside and continue successfully the production process, without loss of profits or time.


  • Detect anomalies on a supply chain (correct product positioning, counting the number of items, surveillance over the assembly line, recognition of the intrusion of unwanted or dangerous objects)
  • Improve product quality through inspection and defect detection (detect even the slightest defect in products based on the standard of quality)

Read about our sports performance analysis technology here 

Related Articles

Gemmo's noise classification case study with Sonitus