Video has emerged as a dominant information tool for everyone from grade school Tik Tokkers to legacy communication companies building their online presence.
But many enterprises that share or archive video may possess imperfect storage and retrieval systems because of the human factor: the time and cost of training, monitoring and evaluating.
A startup that uses machine learning to help evaluate video content has taken a major step forward in its research and development, thanks to its participation in a 5G testing program.
The program is ENCQOR 5G (Evolution of Network Services through a Corridor in Quebec and Ontario for Research and Innovation), which operates five 5G innovation hubs in Ontario and Quebec, with one at Communitech’s Data Hub in Waterloo.
The startup is Vubble, based in Toronto and Waterloo, which helps its clients automate the generation of data, including categorization, thus improving the value of their video assets.
Should a video of a woman breastfeeding her child be categorized as pornography or health advice for mothers?
Vubble’s Video Categorizer automates the labelling of ambiguous and contextual information based on both visual and audio cues and can reduce the reviewing time by staff to mere minutes, saving person-hours and payroll.
Vubble serves such clients as CTV and U.K.’s Channel 4 — with their voluminous archives of video – but the Categorizer will introduce efficiencies to any enterprise that uses and stores video or, for instance, can benefit any agency or company that wishes to monitor mentions of their brands in online videos.
Katie MacGuire, co-founder and Co-CEO of the seven-year-old, six-member company, says she was motivated to apply to use the ENCQOR 5G testbed after attending a Communitech roundtable where Rogers discussed the need for 5G solutions in the future.
“People at the roundtable were all saying that 5G is one of the top priorities, so we figured we needed to prepare for the 5G world and understand what it would mean to our customers.”
Since Vubble’s engineer was Toronto-based, the company used the Toronto testbed, but were guided by the data technology experts at Communitech, as they first established a baseline on what kind of traffic a 5G network would provide, then installed the Categorizer in the cloud to test traffic flow. They ran simulations of clients uploading video to the cloud-based Vubble storage, and then asked the Categorizer to review the footage. Then the simulated client streamed video from Vubble storage. Finally, live video was streamed to Vubble’s storage to see if categorization could happen live. The final test showed that the speed of transmission of the 5G network was a challenge.
MacGuire said the four-month testing “helped us identify development priorities over the next year.” She said she appreciated that the research orientation of the ENCQOR 5G testbed allowed Vubble to do R&D without having to lean on a customer for early testing.
“We’ve done all kinds of different government programs and innovation programs, and this one really was the first one that was purely research-focused. I think the greatest strength of the program was that we had a lot of freedom to try things that were not completely tied to a customer already . . . Our in-house knowledge took a big leap.”
Mauro Rossi, Director of Advanced Technology Platforms at Communitech, says the ENCQOR
5G experience is not limited to the 5G testbed. ENCQOR 5G, through the Ontario Centre of Innovation, covers half the cost of the project, up to $50,000, and although the funded portion is now complete, Vubble has access to the testbed for the foreseeable future, along with access to Communitech’s in-house experts, including an active partnership with a data growth coach.
As well, the learning from the testing benefits both parties: “Our coaches get a lot out of those interactions as well,” Rossi said. “Listening to startups like Vubble helps us to pivot and adjust our programming to ensure these companies get the maximum value that we can deliver.”
MacGuire said that the ENCQOR 5G testbed is not just for idle tinkering. Vubble committed $50,000 and the time of its small team to undertake the tests: “Companies should be aware it takes quite a bit of resourcing. Make sure your customers want it, make sure your experimenting is experimenting that will move you forward.”
But the working environment was ideal: “They gave us all the access we needed. They gave us all the tools we needed. They were super knowledgeable – we came with ‘Here’s what we built’ and they said, ‘Here’s how you’re going to connect it. Here are the different tools. Have you thought about trying it on this type of phone or this kind of laptop?’ They were great.”