Live data from Hacker News

Cloud Video Intelligence API

cloud.google.com

21–30 of 91 posts

Re: Cloud Video Intelligence API

#21
post #3

I'm curious about how much use these general-purpose computer vision APIs are actually getting. How many companies out there really want to sift through a lot of photos to find ones that contain "sailboat"? I'm inclined to think a lot more companies would want to find "one of these five different specific kinds of sailboats performing this action", which is definitely not among the tens of thousands of predefined lab…

I believe it's public that Vision API has already processed more than a billion photos since being made public, which is likely a good sign of usage.

Re: Cloud Video Intelligence API

#23
post #12

I think the most commercially successful application of computer vision has been quality-control devices (citation needed). Agriculture is very interested in CV for a return-optimization technique known as precision farming. Manufacturers pay for inspection of production throughout the pipeline. To predict where a mass-market CV could be successful, I think we should look for industries with similar problems but cann…

I feel like the places where CV always proves most useful are places where either we need more eyes than would be financially viable, or we need eyes in places its hard/unsafe/expensive to have them.

The tipping point on training+pay for a human vs an ML/CV system for a lot of tasks is coming down.

I saw a Microsoft spotlight last fall on a company that's using drones + CV to inspect power infrastructure in Scandinavia. As much as the tech to do that has cost, its probably cheaper than sending out helicopters with specialists all the time to check power line towers for wear/damage.

Re: Cloud Video Intelligence API

#27

Correct me if I'm wrong, but this is just a frame-by-frame labeling. You can download whatever pre-trained CNN, pass individual frames through it and get the same result.

True. But then you have to deploy and maintain that CNN yourself.

The value prop is similar to, say, Twilio. Though, arguably, it's easier to run your own pre-trained CNN than it is to replicate the telephony, VoIP, and video conferencing stuff Twilio provides.

Also, presumably Google is hoping that they can continue to train and improve their CNN so that it's always just a little better than the best free-to-download ones.

Re: Cloud Video Intelligence API

#28
post #12

I think the most commercially successful application of computer vision has been quality-control devices (citation needed). Agriculture is very interested in CV for a return-optimization technique known as precision farming. Manufacturers pay for inspection of production throughout the pipeline. To predict where a mass-market CV could be successful, I think we should look for industries with similar problems but cann…

I feel like the places where CV always proves most useful are places where either we need more eyes than would be financially viable, or we need eyes in places its hard/unsafe/expensive to have them. The tipping point on training+pay for a human vs an ML/CV system for a lot of tasks is coming down. I saw a Microsoft spotlight last fall on a company that's using drones + CV to inspect power infrastructure in Scandinav…

Or we need eyes that can drive a physical response faster & more accurately than a human can react. This is the whole robotics/drone/hardware market.

Right now, most of these efforts work at the same speed humans work (probably because we still need to monitor them to make sure they're doing the right thing). But imagine self-driving cars with no traffic lights, because the cars can react & communicate fast enough to avoid a collision without any macro-scale signaling. Manufacturing processes where molten metal is shaped directly into whatever shape desired, where computers do real-time calculations of the effect of gravity & cooling rate on the position of the material. Airliners that never land and can take on passengers anywhere, because passengers ascend in a personal drone that mates with the mothership under computer control.

Unfortunately CVaaS isn't really suitable for this, because the network delays transmitting the data to a datacenter will outweigh any speed benefits of computerizing it. But I could see a big market for services that let you train a model in the cloud at high-speed, and then download the computed model for execution on a local GPU or TPU.

Re: Cloud Video Intelligence API

#30

Correct me if I'm wrong, but this is just a frame-by-frame labeling. You can download whatever pre-trained CNN, pass individual frames through it and get the same result.

I don't know if they are doing it right now in this API, but Google and plenty of others have demonstrated recognition of actions in videos. I expect they'll add it later if it's not there right now.

At the very least in this release they are identifying scene changes, which is a dynamic property.

Post reply on HN