Overview Shortcuts and Prediction Types
Overview shortcuts for Data, Labels, and Predictions and Prediction Types for Predictions, give you a quick method to filter your unwanted or problematic images. They are shortcuts to improving datasets and model performance. They give you a quick launch pad to improve your data and your model performance.
Data Issue Shortcuts
Data Issues Overview
| Title | Metric Type | Ontology Type |
|---|---|---|
| Duplicates - Duplicate and near-duplicate images. Images with a Uniqueness score of 0.0 to 0.00001 are flagged as duplicates. | image | |
| Blur - Images that are too blurry. Images with a Sharpness score of 0.0 to 0.005 are flagged as blurry. | image | |
| Dark - Images that are too dark. Images with a Brightness score of 0.0 to 0.1 are flagged as too dark. | image | |
| Bright - Images that are too bright. Images with a Brightness score of 0.7 to 1.0 are flagged as too bright. | image |
Label Issue Shortcuts
Label Issues Overview
| Title | Metric Type | Ontology Type |
|---|---|---|
| Aspect Ratio - Identifies objects with a low aspect ratio value. Images with an Aspect Ratio score of 0.0 to 0.1 are flagged as having an issue. | image | bounding box, polygon, rotatable bounding box |
| Border Proximity - Identifies annotations that are too close to image borders. Images with a Border Proximity score of 1 are flagged as being too close to the border. | image | bounding box, point, polygon, polyline, rotatable bounding box, skeleton |
| Low Annotation Quality - Compares image classifications against the 60 most similar images. Images with a Classification Quality score of 0.0 to 0.02 are flagged as having an issue. | image | radio |
| Relative Area - Identifies annotations that are relatively too small compared to the size of the image. Images with a Relative Area score of 0.0 to 0.003 are flagged being too small. | image | bounding box, polygon, rotatable bounding box |
Prediction Issues and Types
Prediction Issues Overview
| Title | Metric Type | Ontology Type |
|---|---|---|
| Shape Outliers - Duplicate and near-duplicate images. Images with a Polygon Shape Anomaly score of 0.0 to 0.02 are flagged as duplicates. | image | |
| Border Proximity - Identifies annotations that are too close to image borders. Images with a Border Proximity score of 1 are flagged as being too close to the border. | image | bounding box, point, polygon, polyline, rotatable bounding box, skeleton |
| Aspect Ratio - Identifies objects with a low aspect ratio value. Images with an Aspect Ratio score of 0.0 to 0.1 are flagged as having an issue. | image | bounding box, polygon, rotatable bounding box |
Prediction Types
| Title | Metric Type | Ontology Type |
|---|---|---|
| All - All model outcomes. | image | |
| All Positives - All model outcomes that are True Positive and False Positive. | image | |
| True Positives - All model outcomes where the model correctly identified objects. | image | |
| False Positives - All model outcomes where the model incorrectly identified objects as the correct object. | image | |
| False Negatives - All model outcomes where the model incorrectly identified objects as the wrong object. | image |
Use Issue and Prediction Type shortcuts
This process assumes that there is already one or more Projects in Active.
To use Issue and Prediction Type shortcuts:
-
Log in to the Encord platform. The landing page for the Encord platform appears.
-
Click Active in the main menu. The landing page for Active appears.
-
Click the Project. The landing page for the Project appears with the Explorer tab selected.
-
Click Data, Labels, or Predictions. The Explorer workspace changes based on what you clicked. The Overview tab displays with the shortcuts.
-
Click a shortcut. A filter is applied to the images. The images appearing in the Explorer workspace changes depending on which shortcut you click.
-
Click Filter if you want to modify the filter settings.
-
Further search, sort, and filter the data.
-
Create a Collection based on the results.
-
Create a Dataset (and Project) and send that Dataset to Annotate.
-
Further annotate your data.
-
Rinse and repeat until you have the dataset you need for optimal model performance.
Was this page helpful?