Training Data or Model

Step 1 is where you provide what the pipeline works from. Start by choosing a workflow, then fill in the inputs that workflow needs.

What This Step Does

• Pick a workflow — train a new model, or import one you already have; the "Choose Workflow" help explains how to decide

• Provide the matching inputs for that workflow using the file selectors below the choice

• Advance when the required inputs are in place: training continues to Step 2 (Configure), import jumps to Step 4 (Inference)

Inputs by Workflow

• Train from scratch: raw training images and their annotation masks

• Use a pretrained model: the model weights (.pth) and its configuration (.json)

Each selector has its own help icon describing the file it expects and how to prepare it.