Links

Training

Once you finish labeling your images, all images will be shifted to the Labeled column.
To submit these annotations for AI training, you need to go to Train/ Deploy and click Train a new AI pipeline.
To submit these annotations for AI training, click Train/ Deploy on the left tab. Here, you will be able to see all versions of AI pipelines that you have trained for this particular project. To train a new pipeline, click the button Train a new AI pipeline.
In the Train a new AI pipeline, you will see the following information.

Description

This is a label for your convenience that will be shown in the AI pipelines table. This will help you distinguish between multiple versions.

Choose classes to include

Select one or more classes that you created to include in training. Any classes you do not check here will be excluded from training. The labels included will be the intersection between the files that were marked as completed, shown in Included images, and the classes that are included.

Create a test set

In this section, you will create a set of data that will not be trained on, and used only to assess performance. This is often called a test set or hold-out set. This is important to realistically evaluate the performance you will get during analysis time. You can read more about training and test sets here.

Errors and warnings

Errors: This section shows any items that are required to be resolved before you can submit. One example of an error is having too few annotations. We require more than 50 annotations for each AI training.
Warnings: Indicates items we strongly recommend you resolve to before submitting the training job for best results. These items will not stop you from submitting the training job.
Here are the rules for Errors and Warnings:
Warning threshold
Error threshold
Number of total labels
Need at least 200 (instance), 50 (semantic)
Need at least 50 (instance), 10 (semantic)
Number of labels for a class
Need at least 75 (instance), 15 (semantic)
Need at least 15 (instance), 3 (semantic)
Number of labels per cropped image
Need at least 1
N/A
Maximum side of largest image
N/A
Cannot exceed 100,000 pixels
Number of Channels per image
N/A
1 or 3 channels
Number of images set aside for testing
Need at least 1 image
N/A
Once you review all Errors and Warnings, you can click continue, which will show you the version, description, and number of credits required to train the model.
Finally, click Start training! Training runs are parallelized across multiple GPUs, and usually finish in 4-5 hours.
You can check training progress from the Train and Deploy screen.
Screenshot with Version 1 of Kidney glomeruli analysis being trained.

Once your model is trained

Once your model has trained, you can evaluate its performance and activate it.

Supported model types

We support the following model types for 2D images, or for z-plane 3D analysis:
  • Instance segmentation (classify, segment, and separate each type of object within an image)
  • Semantic segmentation (classify and segment out object regions within an image)
We use state-of-the-art models for our segmentation models, transfer learning off an optimal dataset, usually based off of Cascade R-CNN and vision transformers.