Image Classification Plugin

This plugin provides the necessary tools to implement image classification tasks using the DashAI framework. The plugin includes a data loader and an MLP (Multi-Layer Perceptron) based classification model.

Components

ImageDataLoader

• Specialized data loader for images
• Compatible with ZIP files containing image folders
• Converts images to bytes format for processing
• Integration with HuggingFace Dataset format

MLPImageClassifier

• Image classification model based on Multi-Layer Perceptron
• Features:
  • Configurable hidden layers
  • Adam optimizer
  • CrossEntropyLoss function
  • GPU support (CUDA)
  • Automatic image transformations (resizing to 30x30)
  • Ability to save and load trained models

Model Parameters

• epochs: Number of training epochs (default: 10)
• learning_rate: Learning rate (default: 0.001)
• hidden_dims: Hidden layer dimensions (default: [128, 64])

Requirements

• DashAI
• PyTorch
• torchvision 0.19.1

Usage

1. Load the image dataset using ImageDataLoader
2. Configure the MLPImageClassifier model with desired parameters
3. Train the model using the fit method
4. Make predictions using the predict method

Notes

This plugin is designed to integrate with the DashAI framework and does not implement a specific model itself, but rather provides the necessary infrastructure for image classification.