🖥
🖥
🖥
🖥
AugeLab Studio Manual
English
Ask or search…
K
Comment on page

Custom CNN Model

Custom CNN Model Node Documentation

The Custom CNN Model node in AugeLab Studio is used to load a pre-trained TensorFlow AI model for image prediction. It allows you to load both TensorFlow (.h5) and TensorFlow Lite (.tflite) models.

Node Overview

The Custom CNN Model node loads a pre-trained TensorFlow AI model for image prediction. It provides the following outputs:
  • Class Number: The predicted class number for the input image.
  • Class Name: The predicted class name for the input image.
  • Raw Output: The raw output of the model for the input image.

Node Properties

  • Node Title: Custom CNN Model
  • Node ID: OP_NODE_LOAD_AI_MODEL

Inputs

The Custom CNN Model node has one input socket:
  1. 1.
    Image: The input image for which the model will predict the class.

Outputs

The Custom CNN Model node has three output sockets:
  1. 1.
    Class Number: The predicted class number for the input image.
  2. 2.
    Class Name: The predicted class name for the input image.
  3. 3.
    Raw Output: The raw output of the model for the input image.

Node Interaction

  1. 1.
    Drag and drop the Custom CNN Model node from the node library onto the canvas in AugeLab Studio.
  2. 2.
    Connect the input image to the node's input socket.
  3. 3.
    Click on the "Load Model" button in the node's properties.
  4. 4.
    Choose the TensorFlow AI model file (.h5 or .tflite) to load.
  5. 5.
    The model will be loaded and the node will display the loaded model information.
  6. 6.
    The node is now ready to make predictions.
  7. 7.
    Connect the output sockets of the Custom CNN Model node to other nodes in the pipeline for further processing or visualization.

Implementation Details

The Custom CNN Model node allows you to load both TensorFlow (.h5) and TensorFlow Lite (.tflite) models. It provides separate loading methods for each type of model.
The node uses the load_model_tf function to load TensorFlow models (.h5 files) and the load_model_tflite function to load TensorFlow Lite models (.tflite files). These functions handle the loading process and set the loaded model in the node's model attribute.
When the node is evaluated, the evalImplementation method is called. It retrieves the input image from the input socket and performs the prediction using the loaded model. The predicted class number, class name, and raw output are returned as outputs.
The Custom CNN Model node supports both TensorFlow and TensorFlow Lite models, providing flexibility in using pre-trained models for image prediction tasks.

Usage

  1. 1.
    Drag and drop the Custom CNN Model node from the node library onto the canvas in AugeLab Studio.
  2. 2.
    Connect the input image to the node's input socket.
  3. 3.
    Click on the "Load Model" button in the node's properties.
  4. 4.
    Choose the TensorFlow AI model file (.h5 or .tflite) to load.
  5. 5.
    The model will be loaded and the node will display the loaded model information.
  6. 6.
    The node is now ready to make predictions.
  7. 7.
    Connect the output sockets of the Custom CNN Model node to other nodes in the pipeline for further processing or visualization.

Notes

  • The Custom CNN Model node is used to load a pre-trained TensorFlow AI model for image prediction.
  • The node supports both TensorFlow (.h5) and TensorFlow Lite (.tflite) model formats.
  • The loaded model can be used to predict the class of an input image.
  • Ensure that the input image is compatible with the loaded model's input requirements.
  • The predicted class number, class name, and raw output can be accessed through the output sockets of the node.
  • The Custom CNN Model node requires the TensorFlow or TensorFlow Lite libraries to be installed.
  • Make sure to select the appropriate model format (.h5 or .tflite) when loading the model.