Reproduce issue with the inference.json

[MelindaDong]

Hi, I tried to reproduce the table 1 result in CVPR2025 paper for the VISTA3D auto, I noticed that you provide the example inference.json file for the spleen task, like this:

" "input_dict": "${'image': '/data/Task09_Spleen/imagesTr/spleen_10.nii.gz', 'label_prompt': [3]}",
"everything_labels": "$list(set([i+1 for i in range(132)]) - set([2,16,18,20,21,23,24,25,26,27,128,129,130,131,132]))","

When trying to do inference with the colon cancer task, I believe the label_prompt should be 27, based on thishttps://github.com/NVIDIA-Medtech/NV-Segment-CTMR/blob/main/NV-Segment-CT/configs/label_dict.json
But not sure how to modify the "everything_labels" part; it seems like you encoded the information that for one specific task, what the available classes are.
Or can you provide the full inference.json file for easy reproduction?

Your timely reply will be very helpful, thanks a lot.

[heyufan1995]

Hi If you want to reproduce the results from the paper, just follow the evaluation part in
https://github.com/Project-MONAI/VISTA/blob/main/vista3d/README_research.md

If you want to do inference, just change "everything_labels" in the inference.json to whatever class you want or
python -m monai.bundle run --config_file configs/inference.json --input_dict "{'image':'example/s0289.nii.gz','label_prompt':[27]}"