Currently, the labels for my object detection model are like: [xmin, ymin, xmax, ymax, class].
Would one hot encoding the classes be beneficial?
Yes! it is important because it helps the data to be rescaled. and as the numeric values are more easy to determine the probability for our values.