yolov5n vs yolov5s

[antimo22]

Hi everybody, I'm using yolov5 models and I was trying to find the differences between yolov5n and yolov5s in terms of architecture, because I would like to cite them in my thesis work. Can anyone help me? The only thing that I'm able to notice is the difference in depth_multiple and width_multiple, but I dont' understand how they affect the model. Thanks in advice.

[glenn-jocher]

@antimo22 these are two model scaling parameters that I adopted from EfficientDet. They are used to scale the model in depth and width. See https://arxiv.org/abs/1911.09070

[antimo22]

Clear. Thank you for your time!

[PraveenMNaik]

Sir, is there any particular case that we need to use one of yolov5(s,n,m,l)... does it depend on the number of images taken?...

[glenn-jocher]

@PraveenMNaik any model can be used on any dataset. Smaller models are faster but less accurate.

[ahmadsakir]

Sir, how do i convert .pt to .ptl or .tflite?

[Heatdh]

You can pass to onnx then export to tflite.
Pytorch (pt) => ONNX => TFlite