As some of the models have their required image input size requirement, do these images have to be resized to that size before generating the tf records, or can i mention it in the training spec file under ‘preprocessing’ which will take care of the desired crop ?
Please see 1.0.1 user guide and 2.0 user guide. They are different.
If you are using 2.0 docker, all detection networks need resizing images/labels.
Note: The tlt-train tool does not support training on images of multiple resolutions, or resizing images during training. All of the images must be resized offline to the final training size and the corresponding bounding boxes must be scaled accordingly.