Skip to content

Image Classification Activity

The Image Classification activity utilizes machine learning models to identify and categorize objects or patterns within an image. By comparing an input image against a pre-trained model, the activity determines the most likely category or label for the content based on a specified confidence level.

Field Description Requirement
Model file name The full file path or identifier of the pre-trained classification model (e.g., .h5 or .onnx). Required
Image file name The file path of the image to be analyzed and classified. Required
Threshold The minimum confidence score (between 0 and 1) required to accept a classification result. Optional
Response variable name The variable where the classification label and confidence score will be stored. Required

Action Types & Examples

Visual Categorization

Analyzing a photograph to determine if it contains a specific document type or object.

  • Format: Threshold: 0.85
  • Example Result: {"label": "Invoice", "confidence": 0.92}

Implementation Examples

Field Setup

  • Model file name: C:\AI\Models\DocumentClassifier.model
  • Image file name: C:\Scans\Input_001.jpg
  • Threshold: 0.75
  • Response variable name: classificationOutput

Execution Parameters

  • Process: The activity loads the DocumentClassifier.model, processes the image, and since the confidence (0.92) exceeds the threshold (0.75), it returns the result to classificationOutput.

Technical Notes

You need to train a model first to use this activity. You can ask for assistance from the support department for training your own model. The performance of the classification is highly dependent on the quality of the training dataset and the image resolution provided in the Image file name field.