YOLOv7: Trainable Bag-of-Freebies • Hacker News Board

Re: YOLOv7: Trainable Bag-of-Freebies

#21

Post by elbigbad » Sun, Jul 17, 2022, 1:40 AM UTC

SrslyJosh wrote: post #14

JustFinishedBSG wrote: post #7

Earlier quoted context omitted.

It's clearly the latter and I don't see why it would be "yikes". Real time detectors are useless if "real time" means 1fps.

What good is speed if the accuracy isn't significantly better than a coin flip? From the paper: > For example, multi-object track- ing [94, 93], autonomous driving [40, 18], robotics [35, 58], medical image analysis [34, 46], etc. LOL, these are all great use cases for a model with < 60% accuracy!

It’s more nuanced than this. It’s not “look at this image and tell me yes or no if there’s a car in it” it’s more like “tell me where all the cars are in this image, if any.” We use this a lot, and ramping up recall we can do some interesting use cases.

I assure you it’s highly useful in the real, real world.

Re: YOLOv7: Trainable Bag-of-Freebies

#22

Post by synergy20 » Sun, Jul 17, 2022, 4:50 AM UTC

montanalow wrote: post #15

squarefoot wrote: post #13

As someone who got only his feet wet with OpenCV like 20 years ago, so basic shape recognition and no AI involved, what read/software, etc. would you suggest to catch up and play with current technology without being inundated by theory that I'm sure I couldn't grasp?

Go to huggingface.com and start with some of the tutorials. The operational/engineering skill sets alone are all you need to treat modern ML models like any other black box API/SDK.

went there and there are lots of stuff indeed, but I failed to find anything related to "operational/engineering skill sets"?

Re: YOLOv7: Trainable Bag-of-Freebies

#23

Post by isoprophlex » Sun, Jul 17, 2022, 4:56 AM UTC

hwers wrote: post #10

isoprophlex wrote: post #4

Github repo mentions "teaser: Yolov7-mask" showing segmentation as well. Highly relevant to my interests. Sadly I can't easily discern any other info on this topic. Anyone knows any more, maybe?

What are you using it for if can share? I’ve thought about training some of these and releasing the weights but I’ve never found a reason they’d really be useful personally so it never really happened

I'm working on a computer vision pipeline that relies heavily on segmentation to detect objects in video feeds. We capture about 6 hours of video each day. So being somewhat close to real time with our processing rate is important ...

Re: YOLOv7: Trainable Bag-of-Freebies

#24

Post by mdda » Sun, Jul 17, 2022, 6:35 AM UTC

squarefoot wrote: post #13

As someone who got only his feet wet with OpenCV like 20 years ago, so basic shape recognition and no AI involved, what read/software, etc. would you suggest to catch up and play with current technology without being inundated by theory that I'm sure I couldn't grasp?

To just play with something : https://huggingface.co/spaces/nateraw/yolov6 (There's an images tab, and some samples below).

If you go to the associated code, you'll see that it needs a 'backbone', 'neck' etc. What is a backbone? Questions that arise directly from the code will lead you towards good blog articles, etc. https://huggingface.co/spaces/nateraw/yolov6/blob/main/yolov...

OTOH, you could go and have a look at (for instance) the Stanford vision courses for a more 'theoretical' approach. But the code itself is often solid guide to what's going on (the frameworks used for Deep Learning map well onto what's being discussed in blogs/lectures/papers).

Re: YOLOv7: Trainable Bag-of-Freebies

#25

Post by DonHopkins » Sun, Jul 17, 2022, 7:24 AM UTC

isoprophlex wrote: post #3

anewpersonality wrote: post #2

We should stop calling it YOLO after the creator quit machine learning.

Especially hilarious considering some other people ALSO jumped on the "we made an object detector so let's call it YOLOvX" wagon and released... Something called YOLOv7. https://github.com/jinfagang/yolov7

Looking forward to the cat detector in YOLOv9.

Re: YOLOv7: Trainable Bag-of-Freebies

#26

Post by DonHopkins » Sun, Jul 17, 2022, 7:26 AM UTC

gchq-7703 wrote: post #8

IncRnd wrote: post #6

In YOLOv7, YOLO and v7 don't go well together. No, not at all. YOLO normally means "You Only Live Once", and v7 means it's lived at least six times before this. While the author likely didn't have that intention, that's what came across. Even for YOLO meaning "You Only Look Once" YOLO and v7 do not go together well.

YOLO in this case stands for "You Only Look One".

YAML originally stood for "Yet Another Markup Language" until somebody pointed out that it wasn't actually a markup language, so they retro-named it "YAML Ain't Markup Language".

Re: YOLOv7: Trainable Bag-of-Freebies

#27

Post by binibus » Sun, Jul 17, 2022, 9:52 AM UTC

anewpersonality wrote: post #2

We should stop calling it YOLO after the creator quit machine learning.

Why? For me at this point YOLO means a family of detectors that in a single pass propose a bounding box per pixel and filters them with some clustering algorithm. When I see YOLOfoo I know what kind of architecture to expect. A more descriptive name like YOLO-tricks instead of YOLOvX would be nice though.

Re: YOLOv7: Trainable Bag-of-Freebies

#28

Post by bj-rn » Sun, Jul 17, 2022, 1:11 PM UTC

squarefoot wrote: post #13

As someone who got only his feet wet with OpenCV like 20 years ago, so basic shape recognition and no AI involved, what read/software, etc. would you suggest to catch up and play with current technology without being inundated by theory that I'm sure I couldn't grasp?

MS put up some courses on github: https://microsoft.github.io/ML-For-Beginners

https://microsoft.github.io/AI-For-Beginners/

Re: YOLOv7: Trainable Bag-of-Freebies

#29

Post by kalenx » Sun, Jul 17, 2022, 1:16 PM UTC

SrslyJosh wrote: post #14

JustFinishedBSG wrote: post #7

Earlier quoted context omitted.

It's clearly the latter and I don't see why it would be "yikes". Real time detectors are useless if "real time" means 1fps.

What good is speed if the accuracy isn't significantly better than a coin flip? From the paper: > For example, multi-object track- ing [94, 93], autonomous driving [40, 18], robotics [35, 58], medical image analysis [34, 46], etc. LOL, these are all great use cases for a model with < 60% accuracy!

The only issue with this comment is that it is _not_ what AP means for object detection... https://www.v7labs.com/blog/mean-average-precision

This is definitely not a coin flip, actually somehow close to what a human would produce, IMHO.

Re: YOLOv7: Trainable Bag-of-Freebies

#30

Post by intpx » Sun, Jul 17, 2022, 1:56 PM UTC

montanalow wrote: post #15

squarefoot wrote: post #13

As someone who got only his feet wet with OpenCV like 20 years ago, so basic shape recognition and no AI involved, what read/software, etc. would you suggest to catch up and play with current technology without being inundated by theory that I'm sure I couldn't grasp?

Go to huggingface.com and start with some of the tutorials. The operational/engineering skill sets alone are all you need to treat modern ML models like any other black box API/SDK.

They call it ‘Tasks’

https://huggingface.co/tasks