API Batch Inference doesn't save labels / Slicer Server Error

[annkapopp]

Hello!

I trained a DeepEdit model from scratch on a custom dataset. Now, I would like to execute the inference over the test images. I tried this by using /batch/infer/ on the API server but somehow, this doesn't work. The /labels/original folder is created but empty and the response body looks like this:

{
"id": "c9ddb697-2a92-44ff-a375-7b32cf5ae1d2",
"status": "SUBMITTED",
"request": {
"model": "deepedit_seg",
"images": "all",
"device": "cuda",
"multi_gpu": true,
"gpus": "all",
"logging": "WARNING",
"save_label": true,
"label_tag": "original",
"max_workers": 1,
"max_batch_size": 1
},
"start_ts": "2023-11-06 12:33:48"
}

The model and the images are found though by executing /datastore/ and /model/info/{model}.
I started the server with:
monailabel start_server --app MONAILABEL/apps/radiology --studies path_to_images --conf models deepedit --conf model_dir path_to_model --conf use_pretrained_model True --conf skip_trainers true

When I try to run inference in Slicer, there is an error "Failed to run inference in MONAI Label Server. Message:: Status 500; Response: Internal Server Error".

I hope that somebody can help me, thank you!

[annkapopp]

Solved it. I had to change the config in /radiology/lib/configs to add the path to my custom trained model.

[diazandr3s]

Hi @annkapopp,

Glad to see you manage to make it work.

Another way of doing batch inference is by running the main file in the radiology app folder: https://github.com/Project-MONAI/MONAILabel/blob/main/sample-apps/radiology/main.py#L267

Just provide these arguments: https://github.com/Project-MONAI/MONAILabel/blob/main/sample-apps/radiology/main.py#L295-L297

This will run MONAI Label without actually staring the server, just as any other python script.

Hope this helps,