Labeling objects

To train an accurate and reliable model, the data in your images and videos must be accurately structured and labeled.

Overview

Use the drawing tools to outline objects or areas of interest in your images and videos. Annotate the outlined data with relevant and informative labels. A model learns that the annotated objects or areas of interest are distinctive and recognizes them during an inspection. Data sets with various detailed information results in reliable and accurate models.  

Label names

To avoid an invalid name error, use the following guidelines when you create label names, class names, and resource names:

Label and class names

  • Only ASCII characters
  • 64 characters or fewer

Resource names

Use only ASCII characters for all resource names including:

  • Data sets
  • Models
  • Project groups
  • Categories
  • Object tags
  • Action tags
  • Data sets
  • Models
  • Project groups
  • Categories
  • Object tags
  • Action tags

Editing label names

When you edit an object label name, all of the data sets that use the previous name are updated to use the new name. However, trained models that use the old object label names to classify or label images are not updated. To use the new label names, you must train a new model.

You cannot rename a category to "Uncategorized". If you want to remove a category from an image so that the image becomes uncategorized, in the Categories section of the filter bar, you must delete the image's category.

Note: You can assign a category only to images. Ensure that no videos are selected.

Data sets for object detection models

When you are preparing a data set for training an object detection model, ensure that the following requirements are met:

  • The data set has at least five images.
  • Every defined object has an object label. Images that do not have object labels are not used to train the model.
Note: When requirements are not met, the model cannot be trained to recognize that object type.

You are training an object detection model to recognize cars, and the data set contains the following parameters:

  • Five images: Ensure that you define and label a car as an object in at least five images.
  • Three images and one video: Ensure that you define and label a car as an object in three images and in at least two frames of the video. Labeling five cars in one image is not adequate.
Note: A data set that has a diverse representation of labeled objects produces a more accurately trained model. The exact number of images and objects cannot be specified, but the number is as high as 1,000 representative images for each class. However, you might not need such a large data set to train a model with satisfactory accuracy.

If your data set does not have many images or sufficient variety for training, use the Augmentation feature to increase the data set.

Data sets for anomaly optimized models

When you are preparing a data set for anomaly optimized model training, ensure that the following requirements are met:

  • Data sets do not contain any anomalous object. Including an anomalous object in a data set reduces the model's ability to identify anomalous objects.
  • All images are high resolution and are taken in similar conditions. For example, the level of lighting or an object's distance from the camera is similar for all images in the data set.

Accuracy can decrease when the number of object classes that are in the training data set increases.