Computer Vision, Machine Learning, Artificial Intelligence; phrases seen often in news articles and trade press, but what do they mean, and how do they impact the world of AV. In this column I will try to demystify the technology and highlight some interesting use cases for education. Plus, I have included some audience polling questions, with a prize on offer too!
What do we mean by Artificial Intelligence? Artificial Intelligence [AI] or more specifically “general AI” is a term used to describe complex machines with the ability to think and reason like human beings; these machines remain firmly in the realm of science fiction, at least for now. What we can do currently is termed “narrow AI”, which means a machine that can perform a specific task as well as, or even better than we humans can. This is where the term Machine Learning comes in – training such a machine with data to enable it to “learn” how to do a certain thing. Machine Learning algorithms are widespread and these can easily be trained with data to recognise patterns; a great application of Machine Learning is in Computer Vision.
Most of the deployed computer vision systems today have been pre-trained to look for specific patterns in a stream of video, and hence have narrow application. Examples of this would be car number plate recognition, object detection (apple, person, cat etc), recognising facial expressions, through to more controversial areas like recognising an individual (e.g. facial recognition). Computer vision systems can excel at recognising the objects they have been trained for. E.g. recognising a certain breed of cat in all lighting conditions, from any angle, with any type of video quality in real-life settings. And with many software frameworks available it is easy to train your own models and create your own customised computer vision system.
Some of the most popular computer vision use cases are:
- In/Out Occupancy Analytics
- People Counting
- Queue Size Detector
- Audience Attention Analytics
- Opt-in Loyalty Program Kiosk
- License Plate Car Identification
- Person of Interest w/ Crowd Blur
- Surveillance Anonymizer Blur
- Indoor Threat Detection
- Large Crowd Analytics
What does this mean for the education? What type of computer vision systems might be deployed, why, and what pros and cons might they bring? Our previous columns in AVNews have highlighted our work on a new hybrid teaching space concept called the Visual Learning Lab [VLL] (can we add links in here to previous columns?); a key part of the VLL is looking at computer vision systems and what value, if any, they can bring to education. Initially the focus is on monitoring HE & FE student sentiment, emotion and engagement but there is potential to expand the scope of work to use facial recognition technology to help create a personalised learning experience for students, as well as support features like badge-less access to buildings and lecture rooms.
THE VISUAL LEARNING LAB
Intel have partnered with UK-based SensingFeeling (https://sensingfeeling.io/ ) to install a computer vision system used to measure student sentiment, emotion and engagement during lectures, as part of the suite of new technologies being trialled in the lab. A key feature of this system is that it is GDPR & privacy compliant; no video data leaves the room, and no individual is recognised or identified. Instead the system uses Intel’s OpenVino™ technology and some smart algorithms to gauge the aggregate engagement levels in the Visual Learning Lab. The technology also captures where each individual has been looking (gaze) and has some additional features like heatmapping and checking social distancing.
