Skip to main content

Model Training

The Model Training page allows users to train machine learning models using uploaded datasets and annotations on the Gesund.ai platform. It supports full configuration of training parameters and provides real-time monitoring for ongoing jobs.

1. Selecting Inputs​

To begin, choose the required inputs for training:

  • Dataset: Select a validated dataset from your data repository.
  • Annotation: Choose the associated annotation project.
  • Labeling Configuration: Pick the correct label mapping setup.
  • Model: Select a supported model architecture (e.g., U-Net, ResNet, MobileNet).

2. Training Parameters​

Configure training-specific parameters:

  • Batch Size
  • Epochs
  • Validation Split
  • Learning Rate
  • Checkpoint Mode (e.g., save best model only)
  • Loss Function
  • Evaluation Metrics
  • Image Dimensions (Height × Width)

These parameters determine how the training process is conducted.

3. Dataset and Model Details​

After selection, details of both the dataset and model are shown, such as:

  • Dataset: Name, source, format, size, and modality.
  • Model: Name, version, type, framework, and associated tags.

4. Starting Training​

Once all inputs and configurations are complete:

  • Review your setup.
  • Click the Start Training button.
  • The training job will begin and appear under the Running Jobs section.

5. Monitoring Training Jobs​

Active and completed training jobs are listed with live updates, showing:

  • Job Name
  • Progress Status
  • Dataset & Annotation Used
  • Problem Type (e.g., Segmentation, Classification)
  • Training Size & Validation Size
  • Epoch Count
  • Backbone Architecture
  • Start and End Times

This streamlined training workflow ensures that models can be configured, monitored, and retrained efficiently — all within a single interface.