Should preprocessing steps like denoising, CLAHE, USM be applied on raw 12bit DICOM x-rays or should one convert 12bit DICOM to 8bit JPEG and apply preprocessing?
Should preprocessing steps be same for training and inferencing workflows?
Hi there, have you tried training and using inference.roboflow.com for model predictions? We take care of the preprocessing steps, etc.
In general, yes, preprocessing steps like Resizing should be applied to the train set and before inference time.