AI-FDU

Project 2: Image Classification Using CNN

Environment Setup:

1. Python Environment: Ensure you have Python installed (preferably Python 3.x).
2. Dependencies: Install required dependencies using pip:

pip install torch torchvision numpy matplotlib tqdm

1. Colab: no need to pip

Code Flow:

1. Model Definition:
   • The code defines a Convolutional Neural Network (CNN) model for image classification.
   • ResNet-18 architecture is used as an example, but you can analyze and discuss other CNN architectures as well.
2. Training and Testing:
   • The code includes functions for training and testing the CNN model.
   • It utilizes datasets provided by PyTorch and DataLoader for efficient data handling.
3. Data Augmentation:
   • You need to design data augmentation techniques to improve accuracy. This may include rotations, flips, or other transformations.
4. Optimizer and Learning Rate Scheduler:
   • Stochastic Gradient Descent (SGD) optimizer with momentum is used.
   • Learning rate scheduling is implemented using ReduceLROnPlateau scheduler.
5. Experimental Requirements:
   • Complete the model training and validation code.
   • Design data augmentation techniques to enhance accuracy.
   • Analyze and discuss the design of other optimizers and learning rate schedules.
   • Analyze and discuss different CNN architectures.

Filling in the Blanks:

1. Model Training Function:
   • Complete the model_training() function to train the CNN model.
   • Fill in the code for optimizer operations, obtaining model outputs, designing and calculating loss, and backpropagation.
2. Model Testing Function:
   • Complete the model_testing() function to test the CNN model.
   • Fill in the code for obtaining model outputs and calculating loss.
3. Data Augmentation:
   • Design image augmentation techniques inside the train_transforms to improve model performance.

Running the Script:

• Execute the script main.py from the command line: python main.py.
• Adjust parameters and configurations as needed for experimentation.
• For colab, just click the “run” button.

Experimental Requirements:

1. Complete Implementation of Image Classification (Model Training and Testing) - 4 points
   • Ensure the code effectively trains and tests the CNN model.
2. Design of Data Augmentation Techniques to Enhance Accuracy - 2 points
   • Implement image augmentation techniques such as rotations, flips, etc., to improve model accuracy.
3. Analysis and Discussion on Optimizer and Learning Rate Schedule Design - 2 points
   • Explore and discuss alternative optimizers and learning rate schedules for training CNN models.
   • Analyze their impact on training convergence and model performance.
4. Analysis and Discussion on Different CNN Architectures - 2 points
   • Analyze and discuss various CNN architectures beyond ResNet-18.
   • Compare their performance, complexity, and suitability for image classification tasks.

Total Score: 10 points

Submission:

Submit the code along with an experimental report (up to 4 pages, No involution) to the e-learning platform.

Note:

• Modify and adjust the arguments, configurations, and experimental setup as necessary.
• Provide detailed explanations in the experimental report, covering code implementation, experimental setup, results, analysis, and discussions based on the specified requirements.