Model EfficientNet

This function block utilizes the EfficientNet architecture for image classification tasks. It allows users to select different versions of the EfficientNet model, specify input size, and choose a pooling method for the model output.

📥 Inputs

Choose Folder 2D Connect any data source that provides input images in the required format.

📤 Outputs

This function block outputs a trained model based on the EfficientNet architecture.

🕹️ Controls

Model Type A dropdown menu allowing users to select from various versions of the EfficientNet model (B0 to B7).

Input Size A field where users can specify the size of the input image. The minimum allowed value is 32.

Pooling A dropdown menu where users can select a pooling method (Maximum, Average, or None).

🎨 Features

Multiple Model Versions Users can choose from different variants of EfficientNet, depending on their computational needs and performance requirements.

Flexible Input Size The input size can be adjusted, allowing the model to be used for various image dimensions.

Configurable Pooling Method Users can choose how the model compresses spatial dimensions in the output, affecting the model behavior and performance.

📝 Usage Instructions

1. Connect Input: Link a source that provides 2D images to the input.

2. Select Model Type: Choose one of the EfficientNet variants from the Model Type dropdown.

3. Set Input Size: Enter the desired input image size in the Input Size field.

4. Choose Pooling Method: Select an appropriate pooling method from the Pooling dropdown.

5. Evaluate the Block: Run the block to prepare the EfficientNet model based on the specified configuration.

📊 Evaluation

When evaluated, this function block outputs a constructed EfficientNet model, which can be used for further training or inference tasks.

💡 Tips and Tricks

Choosing the Right Model

If you're working with limited computational resources, consider using EfficientNetB0 or EfficientNetB1, as they are lighter and require less memory.

Pooling Options

Using Maximum pooling might yield better results in tasks where retaining the highest level of detail is crucial. Experiment with different pooling methods to find what works best for your specific task.

Input Size Considerations

Ensure that your input size is a multiple of 32, as EfficientNet models are optimized for this scaling. A size of 256 is a popular choice for many applications.

🛠️ Troubleshooting

Invalid Input Size

If you receive an error indicating the input size is invalid, ensure that the value entered is an integer and greater than or equal to 32.

Incorrect Image Format

If the model throws an error about image format, verify that you're feeding correctly formatted RGB images, as EfficientNet only supports RGB input.