- 20 Dec 2022
- 7 Minutes to read
Check Label Quality
- Updated on 20 Dec 2022
- 7 Minutes to read
After you train your Model, it's best practice to iterate it by improving your dataset and retraining your Model until you receive the results you want. To improve your dataset, you want to make sure that your images are labeled properly. LandingLens offers a variety of features to help guide users on how to properly label images, like Agreement Based Labelingand the Label Book.
However, what if a user accidentally mislabels some images? Mislabeling images can be costly and can cause issues down the line. For example, let's say you work for a company that produces self-driving cars, and you have a Model to detect objects so the car learns when to stop. If your images are not labeled properly and you deploy your Model, the car may not detect the objects it should, which can lead to accidents and product recalls.
LandingLens offers a feature that can help improve the quality of labels called Check Label Quality. Here's how it works.
After your images are labeled, you can train your Model. The platform will automatically run Check Label Quality on your images. If the platform is very confident that an image's labels can be improved, it will generate suggestions for you (called Label Suggestions).
When LandingLens Suggests New Labels
Images must be labeled in order for LandingLens to display suggestions. For example, if an image is marked as "Nothing to Label", and LandingLens is very confident that there should be a label, then the platform will suggest one. However, if an image is not labeled, then the platform will not run Check Label Quality on that image.
LandingLens will only suggest new labels when the platform is certain that an image's labels can be improved. Here are the scenarios:
- If the platform believes a label is missing.
- If the platform believes an image is not labeled correctly. For example, you labeled an object as a cat, but the platform believes it's a dog.
- If the platform believes the label needs to be adjusted. For example, a bounding box is too large and needs to be smaller.
LandingLens Automatically Runs Check Label Quality
LandingLens will automatically run Check Label Quality if:
- You train a Model.
- You iterate at least one label and retrain a Model.
LandingLens will not automatically run Check Label quality if:
- You retrain your Model without iterating any labels.
- You iterate your labels but do not retrain your Model.
Review Label Suggestions
If LandingLens is very confident that an image's labels can be improved, the platform will display a Label Suggestions banner. To view the Label Suggestions:
- Click the Review button in the banner.
- Review the Label Suggestions that have been generated by hovering over the Accept or Dismiss button. Label Suggestions are represented by dotted lines, while your labels (Ground Truth) are represented by solid lines. Accept or Dismiss the Label Suggestions accordingly. Note:Accepting a Label Suggestion automatically adds the label to your Ground Truth.