Learn how to implement a complete, fully functional, deep learning solution for image classification using Keras, TensorFlow and Python.

Deep Learning

In this project, you will develop and implement a deep learning solution for image classification using the CIFAR-10 dataset.

Objective: Create a fully functional deep learning model capable of accurately classifying input images into one of the 10 predefined categories.

Learning Goal 1: Master the implementation of a complete deep learning solution for image classification using Keras, TensorFlow, and Python.

Learning Goal 2: Familiarize yourself with contemporary datasets, challenges, and issues in computer vision and machine learning.

Learning Goal 3: Learn to leverage deep neural networks through transfer learning.

Learning Goal 4: Demonstrate proficiency in model selection, fine-tuning, and evaluating performance across different approaches to the same problem.

Steps:

  1. Execute "Part 1" of the provided code to ensure the CIFAR-10 dataset is loaded correctly into the notebook.
  2. Run "Part 2" to confirm the dataset`s characteristics: 50,000 training images, 10,000 test/validation images, RGB color images of 32x32 pixels, and a balanced distribution across 10 classes. Visualize random samples from the dataset.
  3. Proceed with "Part 3" to build a baseline CNN classifier from scratch named `model`.
  4. Execute "Part 4" to train the CNN model for 40 epochs. Review the learning curves to monitor model performance.
  5. Implement "Part 5" to record and summarize the test accuracy of `model`.
  6. Analyze the confusion matrix and address the following question: Which classes are most frequently misclassified by the model, and does this align with intuitive expectations?
  7. Continue with "Part 6" up to the end of "Attempt #1" to record and summarize the test accuracy of `model_1`.
  8. Proceed with the remaining steps of "Part 6" for "Attempt #2" and record the test accuracy of `model_4`.
  9. (Optional) Modify the code for "Attempt #2" or create your version ("Attempt #3") and test its performance. If any new model achieves superior accuracy, include it in the summary table.

Question to Address: Despite attempts at transfer learning, both models show disappointing validation/test accuracy. What potential issues could explain this outcome?

  1. Compile your findings into a comprehensive report, including visualizations, code snippets, numerical results, and personal insights gained from the project.

By following these steps, you will gain practical experience in building and evaluating deep learning models for image classification, enhancing your proficiency in Keras, TensorFlow, and Python.