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

YOLOv5: State-of-the-art object detection at 140 FPS

blog.roboflow.ai

Post reply on HN
Print view View original

31–40 of 136 posts

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#31

Post by descenter »

jcims wrote: post #23
ahelwer wrote: post #14

Earlier quoted context omitted.

Not something to brag about. Facial recognition has very few applications outside of total surveillance. We should not respect those who lend it their time and effort.

>Facial recognition has very few applications outside of total surveillance. That's not really for you to decide, is it? You're absolutely free to have that opinion of course. >We should not respect those who lend it their time and effort. Also your choice of course. Facial recognition is essentially a light integration of powerful underlying technologies. Should 'we' ostracize those working on machine learning, comp…

> Facial recognition is essentially a light integration of powerful underlying technologies. Should 'we' ostracize those working on machine learning, computer vision, network and distributed computing, etc?

Couldn't you argue the same way against just about any kind of IED or booby trap? Yet people tend to ostracize those who make them more than they do people who make ball bearings and nails.

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#32

Post by nitrogen »

TehCorwiz wrote: post #29
jcims wrote: post #23

Earlier quoted context omitted.

>Facial recognition has very few applications outside of total surveillance. That's not really for you to decide, is it? You're absolutely free to have that opinion of course. >We should not respect those who lend it their time and effort. Also your choice of course. Facial recognition is essentially a light integration of powerful underlying technologies. Should 'we' ostracize those working on machine learning, comp…

You didn't really address the author's point which was that there don't appear to be compelling uses of facial technology beyond mass automated surveillance. I can't think of other uses and I'd be interested if you can come up with some.

Biometric authentication is one that comes to mind. Facial recognition running locally on my own photo library would also be useful for organizing photos. A cloud-free local-only home automation system that can tell the difference between owners/housemates/guests and customize behavior accordingly would also be nice.

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#33

Post by sterlind »

TehCorwiz wrote: post #29
jcims wrote: post #23

Earlier quoted context omitted.

>Facial recognition has very few applications outside of total surveillance. That's not really for you to decide, is it? You're absolutely free to have that opinion of course. >We should not respect those who lend it their time and effort. Also your choice of course. Facial recognition is essentially a light integration of powerful underlying technologies. Should 'we' ostracize those working on machine learning, comp…

You didn't really address the author's point which was that there don't appear to be compelling uses of facial technology beyond mass automated surveillance. I can't think of other uses and I'd be interested if you can come up with some.

Labeling photos with who's in them, either on social media or in private image albums (creating indexes of people in your photos.)

Arguably this is a front for mass surveillance, or can easily be misused for that, but the ostensible purpose is separate and benign.

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#34

Post by joshvm »

bcatanzaro wrote: post #5

Why benchmark using 32-bit FP on a V100? That means it’s not using tensor cores, which is a shame since they were built for this purpose. There’s no reason not to benchmark using FP16 here.

Not sure about the benchmark, but the code includes the option for mixed precision training via Apex/AMP.

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#35

Post by ericd »

ahelwer wrote: post #14
osipov wrote: post #10

Just recently IBM announced with a loud PR move that the company is getting out of the face recognition business. Guess what? Wall Street doesn't want to keep subsidizing IBM's subpar face recognition technology when open source and Google solutions are pushing the state of the art.

Not something to brag about. Facial recognition has very few applications outside of total surveillance. We should not respect those who lend it their time and effort.

Being able to distinguish between people is pretty foundational to being able to personalize AI applications. If you wanted to make a smart home actually smart and not just full of inconvenient remote controlled appliances, this is pretty necessary.

There are obviously privacy concerns with this example, it’d ideally be fully on-prem.

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#37

Post by knicholes »

sillysaurusx wrote: post #18

We made a site that lets you collaboratively tag a bunch of images, called tagpls.com. For example, users decided to re-tag imagenet for fun: https://twitter.com/theshawwn/status/1262535747975868418 And the tags ended up being hilarious: https://pbs.twimg.com/media/EYXRzDAUwAMjXIG?format=jpg&name=... (I'm particularly fond of https://i.imgur.com/ZMz2yUc.png ) The data is freely available via API: https://www.tagpls.c…

I need a dataset and tags for hair, face, neck, arms, left breast, right breast, nipple, torso. Any tips? I'm training a GAN, but I need to specifically segment the parts, as I don't want nipples in the middle of a face. I don't want to have to manually annotate 1,000 images

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#38

Post by rocauc »

ely-s wrote: post #27

There seems to be an unfair comparison between the various network architectures. The reported speed and accuracy improvements should be taken with a bit of scepticism for two reasons. * This is the first yolo implemented in Pytorch. Pytorch is the fastest ml framework around, so some of YOLOv5's speed improvements may be attributed to the platform it was implemented on rather than actual scientific advances. Previou…

Great points, and hoping Glenn releases a paper to complement performance. We are also planning more rigorous benchmarking nonetheless.

re: PyTorch being a confounding factor for speed - we recompiled YOLOv4 to PyTorch to achieve 50 FPS. Darknet would likely top out around 10 FPS on the same hardware.

EDIT: Alexey, author of YOLOv4, provided benchmarks of YOLOv4 hitting much higher FPS here: https://github.com/AlexeyAB/darknet/issues/5920#issuecomment...

Re: YOLOv5: State-of-the-art object detection at 140 FPS

#39

Post by sk0g »

nitrogen wrote: post #32
TehCorwiz wrote: post #29

Earlier quoted context omitted.

You didn't really address the author's point which was that there don't appear to be compelling uses of facial technology beyond mass automated surveillance. I can't think of other uses and I'd be interested if you can come up with some.

Biometric authentication is one that comes to mind. Facial recognition running locally on my own photo library would also be useful for organizing photos. A cloud-free local-only home automation system that can tell the difference between owners/housemates/guests and customize behavior accordingly would also be nice.

I'm looking into YOLO for this, but it's moreso to verify your selfie == image on document, and we want to avoid sending highly sensitive information to third party providers.

The current service we use, while accurate, costs 50 cents per verification...

Edit: reading through this thread, if the model isn't super massive, we could offer on-browser verification! 27MB is still a hefty download though.

Post reply on HN