Pet Image Classifier

A Python application that uses pre-trained CNN models to classify pet images and determine if they are dogs or not.

Programmer

Name: Ritik Kumar
Date Created: October 3, 2025

Project Description

This project classifies pet images using pre-trained CNN models (ResNet, AlexNet, or VGG) and compares the results against true labels. It also determines whether images correctly classify dogs and dog breeds.

Features

• Supports multiple CNN architectures: ResNet, AlexNet, VGG
• Automatic pet label extraction from filenames
• Dog breed classification
• Statistical analysis of classification results
• Detailed error reporting

Requirements

• Python 3.6 or higher
• PyTorch
• torchvision
• Pillow (PIL)

Installation

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

Usage

Basic usage with default settings (VGG model, pet_images folder):

python check_images.py

Specify a different model:

python check_images.py --arch resnet

Specify a different image directory:

python check_images.py --dir /path/to/images/

Specify a different dog names file:

python check_images.py --dogfile custom_dognames.txt

Command Line Arguments

• --dir: Path to folder of pet images (default: 'pet_images/')
• --arch: CNN model architecture - choices: resnet, alexnet, vgg (default: 'vgg')
• --dogfile: Text file containing valid dog names (default: 'dognames.txt')

Project Structure

.
├── check_images.py                      # Main script
├── get_input_args.py                    # Command line argument parsing
├── get_pet_labels.py                    # Extract labels from filenames
├── classify_images.py                   # Classify images using CNN
├── adjust_results4_isadog.py            # Determine if classification is a dog
├── calculates_results_stats.py          # Calculate statistics
├── print_results.py                     # Print results
├── classifier.py                        # CNN classifier wrapper
├── print_functions_for_lab_checks.py    # Helper functions for checking
├── dognames.txt                         # Valid dog breed names
├── imagenet1000_clsid_to_human.txt      # ImageNet class labels
├── requirements.txt                     # Python dependencies
└── pet_images/                          # Folder containing pet images

File Naming Convention

Image files should follow this naming pattern:

Breed_name_number.jpg

Example: Golden_retriever_05223.jpg

The script extracts the breed name by:

1. Converting filename to lowercase
2. Splitting by underscores
3. Keeping only alphabetic words
4. Joining words with spaces

Output

The script provides:

1. Number of images processed
2. Number of dog images
3. Number of non-dog images
4. Percentage of correct matches
5. Percentage of correct dog classifications
6. Percentage of correct breed classifications
7. Percentage of correct non-dog classifications
8. Lists of incorrect classifications (if any)

Error Handling

The application includes comprehensive error handling for:

• Missing files and directories
• Invalid model names
• Missing dependencies
• File read/write errors
• Classification failures

Notes

• The first run may take longer as it downloads pre-trained model weights
• Ensure all image files are valid and readable
• The dognames.txt file should contain one dog breed per line
• Hidden files (starting with '.') are automatically skipped

Troubleshooting

PyTorch not found:

pip install torch torchvision

PIL/Pillow errors:

pip install --upgrade Pillow

Model download issues: Ensure you have a stable internet connection for the first run to download model weights.

License

See LICENSE file for details.