Cat-Dog-Classifier

Overview

This repository contains a deep learning project aimed at classifying images of cats and dogs. The project includes:

1. A custom Convolutional Neural Network (CNN) built in Pytorch.
2. A fine-tuned ResNet50 model pre-trained on Images Dataset from Kaggle.

Both models have been trained on a dataset imported from Kaggle, and the trained models along with the corresponding Jupyter notebooks are available in this repository.

Table of Contents

• Dataset
• Models
• Installation
• Usage
• Results
• Contributing

Dataset

The dataset used for training and validation is imported from Kaggle. It contains labeled images of cats and dogs.

Models

Custom CNN

A custom Convolutional Neural Network (CNN) was built from scratch and trained on the dataset to classify the images into two categories: cats and dogs.

Fine-tuned ResNet50

A pre-trained ResNet50 model, available in the PyTorch library, was fine-tuned on the dataset. The fine-tuned model is located in the "Finetuning Pretrained model" directory.

Installation

1. Clone the repository:

   git clone https://github.com/Ich-Asadullah/Simple-ANN_cats-and-dogs_classifier_using_pytorch/
   cd Simple-ANN_cats-and-dogs_classifier_using_pytorch

Usage

Custom CNN

Use the Simple-ANN_cats-and-dogs_classifier_using_pytorch.ipynb notebook.

Fine-tuned ResNet50

To train the fine-tuned ResNet50:

Use the finetuning_RsNet50-CNN.ipynb notebook.

Loading Trained Models

To load and use the pre-trained models, refer to the following script:

checkpoint = torch.load('model_path')  # Load the saved checkpoint file
model = MyCNN().to(device)  # Create an instance of the model
# Create a new state dict that excludes unexpected keys
model_state_dict = model.state_dict()
for k in checkpoint['model_state_dict']:
    if k in model_state_dict and model_state_dict[k].shape == checkpoint['model_state_dict'][k].shape:
        model_state_dict[k] = checkpoint['model_state_dict'][k]
# Load the model state dict
model.load_state_dict(model_state_dict)

Results

Custom CNN

• Training Accuracy: 76.18%
• Validation Accuracy: 73.5%

Fine-tuned ResNet50

• Testing Accuracy: 98.5%

Contributing

Contributions are welcome! Please open an issue or submit a pull request if you have any improvements or new features to add.