Participation Quick Start Guide

To start participating in the Challenge:

  1. Pick a task and review the description, particularly the Goal and Data sections:
    1. Task 1: Lesion Segmentation, featuring image contour detection
    2. Task 2: Lesion Attribute Detection, featuring semantic image segmentation
    3. Task 3: Disease Classification, featuring multi-class image classification
  2. Download training data, including input image and response ground truth bundles:
    1. Task 1 training
    2. Task 2 training
    3. Task 3 training
  3. Develop an algorithm for generating responses from generalized input images
    1. As necessary, get support or share your insights on the Challenge Forum
  4. (After July 9th, 2018) (Optionally) Submit your validation results for a preliminary end-to-end check of your system
    1. Download validation input images
    2. Use your algorithm to create predicted responses
    3. Submit these predicted responses to receive an immediate score
    4. This will provide feedback that your predicted responses have the correct data format help check for data issues that may lower your score substantially (e.g. a bug in your script causes every segmentation to be empty)
    5. You may make unlimited serial submissions.
  5. (After July 9th, 2018) Submit your test results
    1. Download test input images
    2. Use your algorithm to create predicted responses.
    3. Submit these predicted responses
    4. Your score will be hidden, but you will receive immediate feedback regarding invalid data formats
    5. You may make unlimited and independent submissions, but only the most recent submission will be used for official judging
  6. Include a PDF manuscript with your submission, describing your algorithm’s approach