YOLOv5: State-of-the-art object detection at 140 FPS
blog.roboflow.ai
YOLOv5: State-of-the-art object detection at 140 FPS
1–10 of 136 posts
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#2EfficientDet 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
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#3Er so this "Ultralytics" consulting firm just borrowed the name YOLO for this model and didn't actually publish their results yet?
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#4Two interesting links from the article:
1. How to train YOLOv5: https://blog.roboflow.ai/how-to-train-yolov5-on-a-custom-dat...
2. Comparing various YOLO versions https://yolov5.com/
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#5Why 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.
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#6What does it take to now use this name?
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#7EfficientDet 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
#8Is 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
#9Er so this "Ultralytics" consulting firm just borrowed the name YOLO for this model and didn't actually publish their results yet?
Yeah, they made the most popular PyTorch implementation of YOLOv3 as well so they're not entering out of the blue, though. https://github.com/ultralytics/yolov3
The author of YOLOv3 quit working on Computer Vision due to ethical concerns. YOLOv4, which built on his work in v3, was released by different authors last month. I'd expect more YOLOvX's from different authors in the future. https://twitter.com/pjreddie/status/1230524770350817280
Re: YOLOv5: State-of-the-art object detection at 140 FPS
#10Just 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.