Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Cloud Video Intelligence API

cloud.google.com

Post reply on HN
Print view View original

71–80 of 91 posts

Re: Cloud Video Intelligence API

#71

Post by catshirt »

chimtim wrote: post #55
catshirt wrote: post #33

Earlier quoted context omitted.

how do you know the implementation details? it would be completely naive to implement it that way, considering there is an entirely new attribute video applies over images which of course is "time". I don't know shit about ML- talking out of my ass here- but I'd be surprised if the algorithms didn't account for changes over time or canonical entity recognition (is this the same boat that was in the last image)?

The linked press release shows an animal is detected -- tiger etc. It does not say tiger running or hunting, which is where the time component would have been used.

the press release says:

> nouns such as “dog,” “flower” or “human” or verbs such as “run,” “swim" or “fly”

that out of the way... i suspect you wouldn't need video to detect those things...

and the screenshot you're referring to is an specific application of the API... not a kitchen sink:

> It can even provide contextual understanding of when those entities appear; for example, searching for “Tiger” would find all precise shots containing tigers across a video collection in Google Cloud Storage.

Re: Cloud Video Intelligence API

#72

Post by microcolonel »

crack-the-code wrote: post #63
tyre wrote: post #50

I think their model should take a second pass on the words and probabilities, independent of the video. Look at their example: Animal: 97.76% Tiger: 90.11% Terrestrial animal: 68.17% So we are 90% sure it is a tiger but only 68% sure it is a land animal? I don't think that makes sense. It could be that this is a weakness of seeding AI data with human inputs. I can believe that 90% of people who saw the video would ag…

It's probably more likely that they want each output to be independent of the other. Certain features may be predominantly associated with a tiger, but not necessarily indicative of a terrestrial animal. If the 9.89% chance that they could have been wrong would have been the case, then that should not influence whether or not it was a terrestrial animal. In my opinion, the consumer of the output values should be able…

It's really tough to take the outputs seriously if they are not even on the same scale. That is, tiger, cat, and bengal tiger all imply terrestrial animal. That means that they should all be scaled around that. That is to say that terrestrial animal would need to be at least max(tiger, cat, bengal tiger).

Re: Cloud Video Intelligence API

#73

Post by sharun »

nostrademons wrote: post #28
madenine wrote: post #23

Earlier quoted context omitted.

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…

For human level vision and speech (and even search) long term we might not need computational clouds and data centers. We don't see them in nature. The models/indexes will likely just come prepackaged on device.

Re: Cloud Video Intelligence API

#75

Post by taneq »

tyre wrote: post #50

I think their model should take a second pass on the words and probabilities, independent of the video. Look at their example: Animal: 97.76% Tiger: 90.11% Terrestrial animal: 68.17% So we are 90% sure it is a tiger but only 68% sure it is a land animal? I don't think that makes sense. It could be that this is a weakness of seeding AI data with human inputs. I can believe that 90% of people who saw the video would ag…

Tigers are 38.57% arboreal.

Re: Cloud Video Intelligence API

#76

Post by visarga »

tyre wrote: post #50

I think their model should take a second pass on the words and probabilities, independent of the video. Look at their example: Animal: 97.76% Tiger: 90.11% Terrestrial animal: 68.17% So we are 90% sure it is a tiger but only 68% sure it is a land animal? I don't think that makes sense. It could be that this is a weakness of seeding AI data with human inputs. I can believe that 90% of people who saw the video would ag…

> So we are 90% sure it is a tiger but only 68% sure it is a land animal? I don't think that makes sense.

Could be a carpet with the image of a tiger, or a toy, or a person in a costume.

Re: Cloud Video Intelligence API

#77

Post by visarga »

wyc wrote: 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…

CV will be the technology that kills what jobs we still have left: in agriculture (picking fruit, weeding), in logistics (picking items from shelves), in custodial work (cleaning bots), security, driving, reading MRIs and x-rays - practically all the jobs that depended on vision and could only be done by people in the last 60 years are going to be automated. When CV is fully deployed, the world will be totally different.

I'm pretty excited about precision agriculture, but for plant-life on earth it will mean that now really there will grow nothing outside the system. Bots are going to monitor all plant life. It might be plant-utopia for some species, with timely water and nutrient dispensing, but for other species it might mean being automatically killed by agrobots.

Re: Cloud Video Intelligence API

#78

Post by visarga »

chipperyman573 wrote: post #51
muzakthings wrote: post #44

Earlier quoted context omitted.

You realize that softmax scores aren't probabilities, right? It's just a relative measure of confidence, scaled such that they all sum to 1.0.

You can't add any of the numbers in the picture to equal 1.0 (or 100)

What you need to do is to take the top prediction and see how accurate it is compared to a test set. The scores on the picture represent confidence not accuracy.

Re: Cloud Video Intelligence API

#79

Post by cs2818 »

tyre wrote: post #50

I think their model should take a second pass on the words and probabilities, independent of the video. Look at their example: Animal: 97.76% Tiger: 90.11% Terrestrial animal: 68.17% So we are 90% sure it is a tiger but only 68% sure it is a land animal? I don't think that makes sense. It could be that this is a weakness of seeding AI data with human inputs. I can believe that 90% of people who saw the video would ag…

It seems like an intersting next step after producing these output labels might be to use something like ConceptNet [0] to evaluate the relationship between the labels and somehow incorporate this as feedback.

[0] http://conceptnet.io

Post reply on HN