YOLOv3 grid question

[doobidoob]

Hi AlexeyAB! Thanks for your support. I have a question about YOLOv3.

Is YOLOv3 able to detect only one object in each grid cell? (I knew YOLO and YOLOv2 can detect only one object in each grid cell.)

[AlexeyAB]

@doobidoob Hi,

Yolo can detect number of objects in each grid cell = number of anchors in this cell.

• Yolo v1 can detect 2 objects in each grid cell (number of anchors is hardcoded 2)
• Yolo v2 can detect 5 objects in each grid cell (number of anchors by default 5 and can be changed)
• Yolo v3 can detect 3 objects in each grid cell X there are 3 grids (13x13, 26x26, 52x52) (number of anchors by default 3 and can be changed)

So default yolov3-spp.cfg model can detect 22743 objects on one image.

https://github.com/AlexeyAB/darknet#how-to-improve-object-detection

for training with a large number of objects in each image, add the parameter max=200 or higher value in the last [yolo]-layer or [region]-layer in your cfg-file (the global maximum number of objects that can be detected by YoloV3 is 0,0615234375(widthheight) where are width and height are parameters from [net] section in cfg-file)

[doobidoob]

@AlexeyAB Thank you very much!