Hey, the size of the video data I want to detect is 640 * 480, so must the size of the data set I prepared be 640 * 480? Or what are the requirements, and what are the requirements for the weight and height of the cfg file during training
This depends on which training frameworks do you use.
In general, you don’t need the size exactly the same as 640x480.
But it’s recommended to have the data with the same aspect ration to avoid any non-natural deformation.