Flower Classification with CNN and Keras Tuner

This notebook shows how to build a Convolutional Neural Network (CNN) for flower classification using TensorFlow and Keras Tuner. It uses the tf_flowers dataset, which contains images of five different flower types: dandelion, daisy, tulips, sunflowers, and roses.

Notebook Structure

The notebook is organized into the following sections:

1. Finding the dataset: Loads the tf_flowers dataset using TensorFlow Datasets and displays some sample images.
2. Initial data processing: Preprocesses the images by resizing and normalizing them. Splits the data into training, validation, and test sets.
3. Writing a simple convolutional network without a tuner: Builds a basic CNN model and trains it on the training data. Evaluates the model's performance on the test set.
4. Finding CNN using tuner: Uses Keras Tuner to find the optimal hyperparameters for the CNN model. This involves defining a search space for the hyperparameters and using a Bayesian Optimization algorithm to explore the search space.
5. TensorBoard: Visualizes the training process and hyperparameter tuning results using TensorBoard.

Requirements

• Python 3.7 or higher
• TensorFlow
• TensorFlow Datasets
• Keras Tuner
• Matplotlib
• NumPy

How to Run

1. Make sure you have all the required libraries installed. You can install them using pip:
2. Open the notebook in Google Colab or Jupyter Notebook.
3. Run the cells in the notebook cell by cell.
4. To view the TensorBoard visualizations, you need to replace /content/tf_flowers/bayesian_opt20241210-195027 with your actual path:

My Tensorboard plots for accuracy and loss (for 10 different Hyperband trials with training and validation).

1. Accuracy

2. Loss