Hi, take these comments with a grain of salt - I only just...

Hi, take these comments with a grain of salt - I only just discovered ICE - and have done nothing more than watch the Japanese videos at this stage. On that note, I can read and speak Japanese - so understanding the videos were not a problem.

I thought it might be worth noting that the majority of the first linked video was focused around the constraints of setting up cameras and blacklist definitions/rules. In contrast it notes how the Icetana system can use deep learning to interpret what the rules should be over time to then report on when something is outside the bounds of the norm.

The use case for deep learning here seems good. I imagine some players in the Japan market could be interested but I also question how well the system really works. Systems like these do tend to struggle outdoors due to lighting and other externalities.

Japan does have some decent expertise in deep learning, one crowd with a strong international reputation are https://preferred.jp/en/
Of the projects that I know PFN have worked on some have been focused on feature detection in videos - although as far as I know the subject matter was sports games statistics dashboards etc.

If the system does perform relatively well, it really does knock it out of the park versus Bosch security solutions offers - or at least to the knowledge of one of the vendors I used in NZ a couple of years ago who is 100% focused in the security camera business (councils/police/railway stations - same sectors noted in the video).

I do wonder what the architecture is like. With a system like this once the deep learning is completed for a location, I imagine you could perform the detection analysis at the edge nodes (camera sites) and reduce your streaming and data storage requirements for recording to a server to only be around periods when an event happens. It would add some complexity in some areas but potentially provide some benefits.

It is definitely an interesting space ripe for some disruption where the solutions are limited and quite expensive for equipment and maintenance while only providing limited added value for customers.