Er so this "Ultralytics" consulting firm just borrowed the name YOLO for this model and didn't actually publish their results yet?
YOLOv5: State-of-the-art object detection at 140 FPS
11–20 of 136 posts
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#12EfficientDet was open sourced March 18 [1], YOLOv4 came out April 23 [2], and now YOLOv5 is out only 48 days later. In our initial look, YOLOv5 is 180% faster, 88% smaller, similarly accurate, and easier to use (native to PyTorch rather thank Darknet) than YOLOv4. [1] https://venturebeat.com/2020/03/18/google-ai-open-sources-ef... [2] https://arxiv.org/abs/2004.10934
Those numbers are quite impressive. YOLOv4 -> YOLOv5 Inference time: 20ms -> 7ms (on P100) Frames per second: 50 -> 140 Size: 244mb -> 27 mb
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#13Is it possible to run these models in the browser, something similar to tensorflow.js?
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#14Just 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.
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#15What does it take to now use this name?
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#16What does it take to now use this name?
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#17Er so this "Ultralytics" consulting firm just borrowed the name YOLO for this model and didn't actually publish their results yet?
The intriguing part is that he has also done research in particle physics (as Ulatrlytics) that has been published in Nature [1].
I had never seen anything like that.
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#18And 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.com/tags/imagenet2012validation.json
It exports the data in yolo format (e.g. it has coordinates in yolo's [0..1] range), so it's straightforward to spit it out to disk and start a yolo training run on it.
Gwern recently used tagpls to train an anime hand detector model: https://www.reddit.com/r/AnimeResearch/comments/gmcdkw/help_...
People seem willing to tag things for free, mostly for the novelty of it.
The NSFW tags ended up being shockingly high quality, especially in certain niches: https://twitter.com/theshawwn/status/1270624312769130498
I don't think we could've paid human labelers to create tags that thorough or accurate.
All the tags for all experiments can be grabbed via https://www.tagpls.com/tags.json, so over time we hope the site will become more and more valuable to the ML community.
tagpls went from 50 users to 2,096 in the past three weeks. The database size also went from 200KB a few weeks ago to 1MB a week ago and 2MB today. I don't know why it's becoming popular, but it seems to be.
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#19Just 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.
Not the the difference matters that much.
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#20Earlier quoted context omitted.
Those numbers are quite impressive. YOLOv4 -> YOLOv5 Inference time: 20ms -> 7ms (on P100) Frames per second: 50 -> 140 Size: 244mb -> 27 mb
f(x)=c, zero size, infinite fps. You should also take some accuracy metric into account ;)