Assignment: Image Classification and Performance

Overview

In this assignment, you will explore image classification in greater depth by comparing different classification models on the Caltech101 dataset. Additionally, you will analyze the memory utilization and performance of two common numerical representations: 32-bit floating-point and 16-bit floating-point.

---

Objectives

1. Understand and compare image classification models on a real-world dataset.
2. Measure classification accuracy across different models.
3. Analyze memory utilization and performance for different numerical representations.

---

Part 1: Learning Resources

Caltech101 Dataset

To learn more about the dataset used in this assignment, refer to the Caltech 101 dataset.

Floating-Point Representation

For an introduction to floating-point representation, read this resource: Floating Point Representation - Basics (GeeksforGeeks).

---

Part 2: Assignment Instructions

1. Open the Colab notebook:
   Image Classification Colab Notebook.
2. Work through the notebook and complete the 10 questions (numbered and bolded).
3. During the assignment, you will:
   • Compare image classification models on the Caltech101 dataset.
   • Measure and report classification accuracy.
   • Analyze memory utilization and performance for 32-bit and 16-bit floating-point representations.

---

Part 3: Submission Instructions

1. After completing the notebook:
   • Export it as a PDF by navigating to File > Print > Adobe PDF in Colab.
   • Save the PDF file with the name image_classification_performance.pdf.
2. Provide a link to your Colab solution in case the PDF does not fully display your work.
3. Commit and push the following to your GitHub Classroom repository:
   • The exported PDF (image_classification_performance.pdf).
   • A text file or README update containing the link to your Colab notebook.

---

Notes and Tips

1. Do Not Procrastinate:
   • Google may limit your GPU usage if you exceed their quota. If this happens:
     • Close all open Colab notebooks and wait up to 24 hours for GPU access to reset.
     • Alternatively, use a secondary Google account to continue your work.
2. Best Practices:
   • Test and compare models thoroughly before submitting.
   • Document your observations and insights in the notebook.

---

Additional Help

If you encounter issues with the dataset or Colab, feel free to:

• Refer to the provided resources.
• Ask for assistance during office hours or on the course discussion forum.

---

Happy coding!