With the Image Labeling feature, you can identify the contents of an image or each frame of live video. Each prediction returns a set of labels as well as a confidence score for each label. Image Labeling can recognize people, places, and things. The underlying ML model was trained on millions of images and hundreds of labels.
If you need to know what objects are in an image, and where they are, consider using Object Detection instead.
Models Compatible with the API
- Your model must be in the TensorFlow Lite (.tflite) or Core ML (.mlmodel) formats.
- iOS Only The name of the input layer must be named
imageand the output
- Android Only The input (image) and output layer (confidence) should be defined in the TensorFlow Lite conversion tool.
- The input should have the following dimensions:
1 (batch_size) x 224 (height) x 224 (width) x 3 (num_channels). Height and width are configurable.
- The output should have the following dimensions:
1 x number_of_labels.
Architecture Format(s) Model Size Input Output Benchmarks MobileNet V2 variant Core ML (iOS), TensorFlow Lite (Android) ~13MB 224x224-pixel image Label + confidence score (0-100%) 38 FPS on iPhone X, 10 FPS on Pixel 2
- The image labeling model supports 1,000 labels. View the full label list.
Customizing Models for Image Labeling
If you have your own dataset and would like to train a custom model that is compatible with the Image Labeling API, sign up for the Standard Plan on Fritz to access training notebooks.