Compare YOLOv4 and YOLOv4-tiny

Compare YOLOv4, YOLOv4-tiny, and two other variations. Minimal training, 4 classes, 100 annotated images, and max_batches set to 10K for the neural network.
Darknet/YOLO discord: / discord

[@AdicttedGamer]

Excellent video. Thank you Stephane.

[@ninhnguyenquang3731]

Hey St, i have a question. Could you help me? Tiling technique is used in training or detecting phase?

[@StephaneCharette]

Both: www.ccoderun.ca/darkmark/ImageSize.html and www.ccoderun.ca/darkhelp/api/Tiling.html

[@StephaneCharette]

Which you can see in the video starting at 1:12

[@ninhnguyenquang3731]

@@StephaneCharette but tiling in detection may get none object when it come to edge of tiles. See 1:58. I read document and see that you devide big img to 4 tiles?

[@StephaneCharette]

@@ninhnguyenquang3731 No, in this case, the image is divided into 15 tiles (5x3) which I briefly mention at 1:51. See my other videos for more information on how tiling works. For example, th-cam.com/video/Dp303GuxjFs/w-d-xo.html and th-cam.com/video/Oz-49MpO2rQ/w-d-xo.html

[@baconga3844]

@@StephaneCharette I am Ninh Nguyen Quang. ( My litle daughter get my phone ^ ^).
Thank you very much for showing me and for this video.
N.

[@rohanpokale4651]

Hello sir, I liked your video , and had some questions, we are also working on a project and planning to use Tiny Yolo as it has faster response and can work on low end systems. So first of all I wanted to know what is distance between camera and the extreme end of ground shown in the video. And also Can you please share link of tutorial video for training Tiny Yolo. Thank You.

[@StephaneCharette]

I have no idea what the distance is. I was standing at the top of the hill by the school looking at my kids playing on the field. As for tutorials, please see my channel. I have lots of tutorials.