Compare YOLOv4-tiny and YOLOv7-tiny

Compare YOLOv4-tiny and YOLOv7-tiny using a custom dataset.
Please watch my other tutorial videos if you are looking to create neural networks like this. For example, start here: • Building and using Dar...
Darknet/YOLO discord server: github.com/AlexeyAB/darknet

[@roger10werneck]

Thanks for the video.
You did a great job in comparing the different versions of YOLO
This is something I've been waiting for

[@glowish1993]

great video, appreciate it a lot and its tough to find objective comparisons of the two

[@nailsonlandim]

Thanks Stephane!

[@Shadow-gc1em]

hey how I train yolov7 tiny? help pls

[@patimuramm9051]

stephane, i already using yolov4-tiny and yolov7-tiny. this result is yolov4-tiny have a better mAP graph then yolov7-tiny. how can?

[@austing.8682]

Darn was hoping yolov7 would be faster, I have a very specific use case that requires minimum latency to work well. Just curious is the code for this card detection available?

[@StephaneCharette]

What code? This is a neural network I trained. There is no code. (Other than Darknet of course.) See the FAQ if you are worried about FPS and latency: www.ccoderun.ca/programming/darknet_faq/#fps

[@XAE-L-xy7mi]

There is a problem here, I got different results

[@StephaneCharette]

Different results of what? Where? How?

[@snim9515]

How is this even possible?

[@StephaneCharette]

How is what possible?

[@snim9515]

@@StephaneCharette v7 is worse than V4? They tested it cifar100 and v7 was better. Why is it not better in your case?

[@StephaneCharette]

@@snim9515 Because I'm not using overly simplistic 32x32 images from CIFAR, or carefully-curated MSCOCO situations. I'm using real world images, real-world datasets, not networks trained to look good in only 1 specific situation.