There is a new wave of cognitive services based on video and image analytics, leveraging the latest in machine learning and deep learning. In a recent SNIA Cloud Storage Technologies Initiative (CSTI) webcast â€œHow Video Analytics is Changing the Way We Store Video,â€ we looked at some of the benefits and factors driving this adoption, as well as explored compelling projects and required components for a successful video-based cognitive service. This included some great work being done in the open source community.
In the course of the presentation, there were several discussion points and questions that arose. Our SNIA presenters, Glyn Bowden from HPE and Kevin Cone from Intel provide the answers.
Q: The material several times mentioned an open source project for video analytics. Is that available for everyone to view and contribute?
A. Absolutely.Â The Open Visual Cloud Product (OVCP) is located on GitHub at: https://github.com/OpenVisualCloud. Contributions are welcome, and there are a significant number of contributors already involved in the project. There were several examples on the versatility of OVCP, and it was noted how extensible the project could be with addition of new tools and models.
Q. Glyn talked about some old video platforms. Did people really capture video like that? Did the dinosaurs roam the tape archives back in those days?
A. Ha! Glyn would like everyone to know that while that was the way video used to be captured, he was not around during the time of the dinosaurs.
Q. Is there a reason to digitize and store old video of such poor quality?
A. Glyn demonstrated how much of this video can still be valuable, but he also discussed how it was difficult to capture and index. Clearly, there are significant storage implications in digitizing too much old video, though cloud storage certainly provides a variety of solutions.
Q. There was a good example of video analytics in smart cities. Is there a role for computational storage in this type of application?
A. Not only is there a role for computational storage, thereâ€™s a significant need for smart networking. Kevin and Glyn provided some cases where the network might do local analytics. In fact, there was a recent SNIA webcast on â€œCompute Everywhere: How Storage and Networking Expand the Compute Continuumâ€ that discussed some aspects of the edge and cloud interaction.
Q. There was a good discussion on governance of video data. One discussion point was around the use of video in public safety and law enforcement. Would it be the case that smart city video might also be useful as a legal tool, and would have different retention rules as a result? Are there other examples of something like this?
A. There are a variety of rules on archiving and retention of data that may be used in public safety. This is a pretty fluid area. Another example would be of videos where children are present, as there are significant privacy issues. The EU leads in the legislative efforts in this area, and they have a number of rules & guidelines that are outlined here.
Q. Digital camera pickups have the ability to see beyond the human visual spectrum. Are there uses for video analytics in the IR and UV spaces?
A. Kevin mentioned the use of IR as an indicator of remote temperature monitoring. Glyn said that this might also be an example of legal hazards, where there could be a violation of health protections. So, governance is likely to play a role in this area.
Q. What are some differences between analytics and storage of video at the edge and storage in the cloud or data center?
A. Video storage at the edge is likely a temporary thing. It could be stored there or analyzed there to reduce the latency of decision-making. Or itâ€™s possible that it would be analyzed to determine how much of the video should be permanently archived. Cloud storage is more permanent, and analytics in the cloud is more likely to generate metadata that would be used to make policy at the edge.