Do we have do do image resizing to train and test using R-CNN object detector in matlab?

I am using R CNN object detector to detect other object. But getting wrong bounding boxes for many cases. Do we have to resize the images or any image size is ok? I have not resized when I used image labeler app for labelling the images also.

Answer 1

[VaL V6]

Hello,

yes, you have to resize both the images and the boudig box that you want to use to train the detector. You can perform this process anytime.

There is a detailed explanation here:

https://it.mathworks.com/help/vision/examples/object-detection-using-faster-r-cnn-deep-learning.html