CS6910_Assignment_-2

This repository contains the Python files used to generate the results presented in the accompanying report. For detailed insights, please refer to the report available here.

Instruction to run the py file:

Usage

train_partA

To run the train_partA.py file, you can use the following command-line arguments:

python train_partA.py -e <epochs> -b <batch_size> -lr <learning_rate> -a <activation> -nf <num_filters> -ks <kernel_size> -dp <dropout_prob> -nd <neuron_dense> -fo <filter_org> -datao <data_aug> -bn <batch_norm> -train <train_folder> -test <test_folder>

Command-line Arguments

• -e, --epochs: Number of epochs to train the neural network (default: 5).
• -b, --batch_size: Batch size used to train the neural network (default: 32).
• -lr, --learning_rate: Learning rate (default: 0.001).
• -a, --activation: Activation function (default: 'relu'). Choices: "relu", "gelu", "mish".
• -nf, --num_filters: Number of filters (default: 32).
• -ks, --kernel_size: Kernel size (default: 5).
• -dp, --dropout_prob: Dropout probability (default: 0).
• -nd, --neuron_dense: Number of neurons on the dense layer (default: 50).
• -fo, --filter_org: Filter organization (default: 'same'). Choices: "same", "halve", "double".
• -datao, --data_aug: Data augmentation (default: False). Choices: True, False.
• -bn, --batch_norm: Batch normalization (default: False). Choices: True, False.
• -train, --train_folder: Directory path for the training dataset.
• -test, --test_folder: Directory path for the test dataset.

Note: Replace <epochs>, <batch_size>, <learning_rate>, and other placeholders with appropriate values.

train_partB

To run the train_partB.py file, you can use the following command-line arguments:

python train_partB.py -e <epochs> -b <batch_size> -lr <learning_rate> -train <train_folder> -test <test_folder>

Command-line Arguments

• -e, --epochs: Number of epochs to train the neural network (default: 5).
• -b, --batch_size: Batch size used to train the neural network (default: 32).
• -lr, --learning_rate: Learning rate (default: 0.001).
• -train, --train_folder: Directory path for the training dataset.
• -test, --test_folder: Directory path for the test dataset.

Note: Replace <epochs>, <batch_size>, <learning_rate>, and other placeholders with appropriate values.

Content overview:

The repository is divided into two parts Part A and Part B and jupyter notebooks are given for the respective questions of the assignment.

Part A

1. Question_1:

   • This notebook is dedicated to creating a Convolutional Neural Network (CNN) model with the following architecture: five convolutional layers, each followed by activation and max-pooling, a dense layer, and finally, an output layer with ten classes.

2. Question_2:

   • This notebook demonstrates the process of running a hyperparameter sweep to explore specific configurations and optimize model performance.

3. Question_4:

   • The notebook executes the model with the best hyperparameter configuration obtained from the sweep and reports the accuracy for the test dataset. Additionally, it generates a plot to visualize the model's predictions and compares them with the original labels.

Part B

4. Question_3:
   • The notebook demonstrates the implementation of a pre-trained model, which is fine-tuned by freezing all layers except the last one and then training only the last layer. Additionally, it reports the accuracy achieved on the test dataset.