Closed FaresAlMohamad closed 10 months ago
Thanks for posting this issue, the enhancements will be helpful. Welcome to contribute the updates, and if you meet any issues when doing the changes, feel free to ask.
This feature is implemented by @FaresAlMohamad in #482
This problem is about the inference.json config file for the prostate_mri_anatomy model. The transformations applied to the test data during preprocessing are not reversed during postprocessing; thus, the resulting predictions have a different rotation compared to the original images and labels. I would like to add an invertd() function to postprocessing.
Additionally, running the inference on a PC without a GPU leads to an error because the handlers function 'CheckpointLoader' attempts to load the model to the GPU. I want to add a parameter that loads the model to the CPU when the GPU is unavailable.