Inference on pre-trained models?

Can you please provide us details on how datalist.json file look for doing inference.
I tried with below format on classification model

{
"label_format": [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1],
"training":[],
"validation": [
	{
		"image" : "1.png"
	},
        ]
}

I am getting below error

Traceback (most recent call last):
  File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "common/inference/scripts/inference.py", line 189, in <module>
  File "common/inference/scripts/inference.py", line 123, in main
  File "common/inference/prediction_utils.py", line 442, in create_inference_file_list
  File "classification/data/utils.py", line 103, in load_dataset_classification
KeyError: 'label'

The data specification json files included in the container at /opt/nvidia/medical/segmentation/examples/brats/ should serve as good examples.

If those examples don’t help solve your issue, could you detail what tlt command you’re running and how so we understand what you’re trying to run?

Thank you @ryanlee, I was able to do with the help of provided examples.
Can we do segemntation on 2D images using tlt-train command?

Hi Akhila,
Sorry but currently we do not support using tlt-train for segmentation on 2D images

Hi

I have 2 questions

  1. What is the command to make predictions on the testing (non-labeled) set?

  2. Does the infer.sh read the label from the validation set to predict the output?