How to resize images and labels for custom datasets?

I use TLT container of v2.0_dp_py2 for training faster rcnn model in my dataset.

In started guide doc, I see “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.”

My dataset images are different resolutions, including 1920x1080, 1280x960, 1280x720 and so on, and I have labeled them with voc format xml file.My question is whether have tool for resize them?

And if I train a modle for a resolution like 1920x1280, whether the model can only detect the image for 1920x1280? How about other resolutions?


Please ignore the question about dataset, I have achieved resize image and label.

How about the infer for different resolutions? I would deploy it in deepstream, can I detect different image/video with tlt model?


Yes, you can deploy it in deepstream. It can detect different resolution videos.
You can also run tlt-infer against different resolution images.