AIAA server response error in MITK workbench

Healthcare Clara Train: AI-Assisted Annotation SDK
#1

I have uploaded model in the MITK workbench from API visualization. The server is also showing the uploaded model details. I have also entered the ‘server URL’ in MITK preference but when I am trying to do ‘Nvidia Segmentation’ the following error is displayed.

ERROR:

Failed to communicate with Nvidia AIAA Server
Fix server URI in Nvidia AIAA preferences (Ctrl+P)

nvidia.aiaa.error.102
description: Failed to parse AIAA Server Response

Kindly response!

#2

Hi
Thanks for your interest in clara Train AIAA.

To debug your issue:

  1. splease open a browser check 
    http://<AIAA_ip>:5000/v1/models

    you should see a list of models

  2. You should go to preference under Nvidia AIAA set path to 
    http://<AIAA_ip>:5000/v1

Hope that helps.

#3

Thank you for your valuable response!

The current problem which I am facing after the execution of previously mentioned reply is that after clicking the button ‘Auto Segmentation’ in the 3D Tools dialog box is as follows:

ERROR:
An error occurred. You should save all data and quit the program to prevent possible data loss.
Dimension 4 is not in (2)(3)

The dimensions of my data is 4x240x240x155 (Brats 2017).

#4

4D inference is not supported in the MitkWorkbench plugin yet. However, the python and c++ clients support it: https://docs.nvidia.com/clara/sdk-api/docs/index.html

You can also use the browser to try it out, i.e. going to http://<AIAA_ip>:5000/docs
and select “POST” /v1/segmentation and then the “Try it out” button.

#5

Hi，
How to get the Nvidia AI-Assisted Annotation/Server URI in MITK workbench?

#6

Hey!
In Clara-Train-SDK, after running the image (step-10) and starting the annotation server command (step-11):

open the link “localhost:5000” in the browser to start the Nvidia AI-Assisted Annotation Server.

I hope it helps :)

#7

For that please go to preference menu in MITK workbench and set url as “http://localhost:5000/v1” and for more information refer the 2nd message in this conversation above.

Above server assumed as localhost.

#8

hello,
I also have some questions,it seems to be kind of blackbox for me and can you explain for me?

my want to use Nvidia autosegment in MITK, And I installed “2018.04.2 with Nvidia AI-Assisted Annotation v1.0.1” & “NvidiaAIAAClient-1.0.1-win64.exe”.

1)Did not the clara-train-SDK interagted in this MITK version? if not, how to install this clara-train-SDK in win10
2)Followed https://docs.nvidia.com/clara/aiaa/sdk-api/docs/quickstart.html, i install the
NvidiaAIAAClient, but when runing “example,cpp”, error occured:“14:45:50 [ERROR] [curlutils.cpp:83 - doGet()] Host
not found: my-aiaa-server.com nvidia::aiaa::exception => nvidia.aiaa.error.101; description: Failed to communicate
to AIAA Server”
what should I do to fix the error
3)I don’t konw how start the AIAA client and how to connect with MITK

Would be helpful if you could share or direct me to any resource where i can get a full path of this end to end workflow with an example.

#9

Hi, please follow the instructions here on how to start the AIAA server: https://docs.nvidia.com/clara/aiaa/tlt-mi-ai-an-sdk-getting-started/index.html

Once that is running, you can use MITKworkbench to communicate to it. Also, see this video for a quick overview of the SDK: https://youtu.be/T0Pjki4vXx0

#10

Hi,

To start the AAAI server, does it require Ubuntu? Can I do this on Windows?

#11

Currently, we have these software Requirements:
Ubuntu 16.04 LTS
NVIDIA GPU driver v410.xx or above
nvidia-docker 2.0 installed, instructions: https://github.com/NVIDIA/nvidia-docker.

Please see https://docs.nvidia.com/clara/aiaa/tlt-mi-ai-an-sdk-getting-started/index.html#requirements for more information.