Understanding the Models Page

The Models page displays all your training runs and their performance metrics. Use it to track model quality, compare configurations, and manage model documentation.

Page Overview

The page shows:

• Statistics cards - Total models, average mAP50, and best mAP50
• Tag filter - Filter the list to show only models for a specific tag
• Models table - All training runs with key metrics

Key Metrics

Metric	What It Means
Test mAP50	Model accuracy score (higher is better, max 100%)
VQC Score	Visual quality rating from manual review (available after VQC is completed)
No Intervention %	Percentage of predictions needing no correction

Viewing Model Details

Click any row to open the model detail dialog with six tabs:

Tab	Contents
Overview	Key metrics at a glance, dataset summary, and training info
Metrics	Detailed Test/Train metrics (mAP50, mAP50-95, Precision, Recall) and VQC metrics
Dataset	Image counts, class breakdown, and annotation process split
Config	Training configuration (structured or JSON view, can copy/download)
Metadata	Model ID, tag, training times, and duration
Notes	Model documentation and observations

Editable Features

You can edit the following information for each model:

Notes

In the Notes tab, click Edit to add or update notes about the model using the rich text editor. Use notes to record training observations, issues encountered, or next steps.

Annotation Process Split

In the Dataset tab, you can edit the annotation process split to control which annotation processes are used for training vs. testing. Click the edit button next to the annotation split section to modify the configuration.

Start Training Time

In the Metadata tab, you can edit the training start time if it needs to be corrected. Click the edit icon next to the time value to modify it.

Comparing Models

To compare training configurations between models:

1. Select models using the checkboxes in the first column
2. Click Compare Configs in the selection banner that appears

Tip: Select models from the same tag to see how different training configurations affected performance.