Machine Learning part of Fasimagiland

Description

Data:

• Utilizing the Google QuickDraw! dataset.
• This dataset contains 43 classes with a total of 258,000 data points.

Augmentation:

• Performing data augmentation using various techniques such as:
  • Rotating
  • Shifting
  • Shearing
  • Zooming
  • Flipping

Preprocess:

• Conducting data preprocessing with steps such as:
  • Reshaping
  • Normalizing
  • Compressing

Model:

• Using the pre-trained ResNet50 model from TensorFlow.
• Adding additional MLP (Multilayer Perceptron) classifiers.

Loss and Optimizer:

• Using Cross Entropy Loss as the loss function.
• Using Adam Optimizer as the optimizer.