Skip to content

QUICK_START.md

Latest commit

 

History

History
92 lines (61 loc) · 3.32 KB

QUICK_START.md

File metadata and controls

92 lines (61 loc) · 3.32 KB

ImageJ/Fiji Plugin

  • Download Fiji (version >= 2.3.0)

  • For ready-to-use ImPartial plugin, user can get a pre-compiled .jar file here. Copy the .jar file into Fiji's plugins directory. For example, if you're using macOS

cp ~/Downloads/impartial_imagej-0.1.jar /Applications/Fiji.app/plugins
  • Restart Fiji
  • From Plugins menu bar, open ImPartial
  • Or Search ImPartial in Search bar as shown below and click Run plugin

No-Code Cloud Execution

User can request our cloud deployed MONAI server to readily annotate and segment the data without needing to compile or run any code locally.

Plugin Features

MONAI Setup

Local setup

  • Start the monai server: monailabel start_server -a api -s /dataset/
  • http://localhost:8000
  • All the data with the labels are saved at the “/dataset/” location given by the user when starting the server.

GPU server request

  • Signup with credentials

    signup

  • Login with your user ID

    login

User Session

  • User session is saved for each user with date and time as the id. User has access to the previous sessions.

  • Each user session has a window of 2 hours at each login with a 5 min warning before the session ends.

  • Restore session: User can restore the session by login-in with their credentials and clicking “select”

    session

Dataset images

  • Click “ADD” to upload your images.
  • Impartial supports .png and .tiff file formats
  • Multichannel images upto 3 channels
  • Example dataset:
    • VECTRA_2CH
    • images are in .tiff and labels (ROIs) are in .zip format

Creating Labels

  • Use the ‘selection brush tool’ from the Fiji toolbar

    selection-brush-tool

  • After creating each label add each label to the Roi Manager add label option.

  • After all the labels are added for the image, press submit to upload the labels on the server. (Warning: do not switch to another image before submitting, otherwise your labels will not be saved)

    add-Label

  • User can also upload previously created/saved labels eg. “roi.zip” by clicking File > open > imageLabel.zip

Upload pre-trained model

  • User can upload a pre-trained model for inference and/or fine-tuning.
  • Example vectra_2ch.pt

Infer

  • Press Infer to view the predictions. User can also change the thresholding to fine tune the results in real time.

Save model

  • User can download the trained model for future training/inference.

Training

  • Training Hyperparameters: User can set the training hyperparameters according to the dataset and requirements. We recommend:
training:
  epochs: 200
  patches: 4000
  patience: 50
  • Click 'start' to begin training.
  • User can see the progress of the training of number of epochs completed on Fiji toolbar.