Training a Darknet/YOLO neural network in less than 90 seconds
ฝัง
- เผยแพร่เมื่อ 28 เม.ย. 2024
- Training a Darknet/YOLO neural network in less than 90 seconds using only 24 images.
- Darknet: github.com/hank-ai/darknet
- DarkHelp: github.com/stephanecharette/D...
- DarkMark: github.com/stephanecharette/D...
- Darknet/YOLO discord server: / discord - วิทยาศาสตร์และเทคโนโลยี
All of the files (images, videos, annotations, cfg, weights) are available for download here: www.ccoderun.ca/programming/2024-05-01_LegoGears/
thanks for the explanations @Stephane Charette!👏👏
Esse cara domina, me ajudou muito quando eu estudava reconhecimento , obrigado Stephane.
Ele é gente boa mesmo, fez algum projeto mano? Trabalha com isso ou só curiosidade msm?
@@leo1722467 fiz mano tá nos meus vídeos reconhecimento de armazenamento de fogo, mais depois eu parei, eu tava ou tô ainda até hoje no grupo dele no Discord,aí vi eles comentar sobre uma técnica para girar o retângulo de reconhecimento era augo novo que ainda tava lançando, aí na época eu parei , esperando essas coisas aparecer e não voltei até hoje.
Hello, thanks for the video. How would this perform on a jetson nano if possible at all given its older version of ubuntu
See this post: www.ccoderun.ca/programming/2021-10-16_darknet_fps/
What's your hardware configuration?
My GPU is a NVIDIA GeForce RTX 3090.
@@StephaneCharette CPU, RAM?
@@ckpioo Ubuntu 20.04, 24 GiB vram, 32 GB ram, AMD Ryzen 9 @ 2200 MHz. Doesn't matter, the point is not to actually train in 85 seconds, the point was to show that if your network is limited to a specific situation, you can train in a very short amount of time instead of "days" and with thousands of images.