123 points by hshen 6 months ago flag hide 19 comments
lucidjour 6 months ago next
Have to agree. YOLO v4 completely changinh the ball game for object detection.
johnsmith 6 months ago prev next
Great article! I've been following YOLO versions and this one seems very promising.
geekytech 6 months ago next
Agreed. YOLO v4 has shown impressive results in object detection. I would love to try it out.
randomuser 6 months ago prev next
Anyone tried it with TensorFlow? Can't wait for an official release.
gabrielw 6 months ago next
Yes, you can convert the weights from Darknet format to TensorFlow. I've done it and it worked fine.
davies_123 6 months ago prev next
I'm excited to see it in real-world applications like autonomous vehicles.
anonymouscow 6 months ago next
Definitely, it could significantly improve the vision capabilities of self-driving cars.
appleseed 6 months ago prev next
Are there any new features from the previous versions?
xzx 6 months ago next
Yes, YOLO v4 introduces Bag of Freebies and Bag of Specials, among other improvements.
alltheintegrals 6 months ago prev next
I read about the Paper and it seems like it's pushing the limits of object detection.
andrew_00 6 months ago next
It certainly does. Exciting to see the advancements in this area.
lemoncode34 6 months ago prev next
Confidence scores are much better in YOLO v4. Impressive work!
tmtms 6 months ago next
By how much? Can't wait to run a comparison with other models.
cvastern 6 months ago prev next
I'm hoping to see updated implementation of YOLO in the OpenCV library.
hrishibr 6 months ago prev next
Would love to play around with the dataset used in the study.
tanmoyin 6 months ago prev next
Can we expect the C++ version from the developers soon? I'm itching to test it.
typical101 6 months ago next
The GitHub repository contains the Darknet code. You could try implementing it in C++ yourself.
moncid 6 months ago prev next
Thenew version is claiminh 58 FPS at Full HD. That's bonkers.
karenso13 6 months ago next
Impressive indeed, got to try it out ourselves.