How to customize or reduce the number of classes in yolov4 model?

There are 80 number of classes are there in the yolov4 - coco matlab model. But, i dont require the all those clasess. With the same model, is any possible to call the particular class while doing the testing? Like, I need only 3 classes such as Car, truck, Person - In that case, is any possible to extract those 3 class trained data alone? If yes, please let me know.

or

Is anyone has the yolov2 pretrained model for person detection along with vehicle? (Totally, 2 classes)

Answer 1

[Sanjana]

Hi Jesse,

I understand that you are looking for a way to customize the “yolov4ObjectDetector” by reducing the no. of classes, which are equal to 80 for a standard “yolov4ObjectDetector”.

Transfer learning can be applied on the pre-trained “yolov4ObjectDetector”, to detect any number of classes.

Please refer to the following documentation, for creating a custom “yolov4ObjectDetector”,

https://www.mathworks.com/help/vision/ref/yolov4objectdetector.html#mw_89976544-2f97-4677-9268-d2d6f8918fed

Hope this helps!

Regards,

Sanjana