Cat and Dog Breed Classification Using ResNet50

---

This project implements a deep learning model for classifying images of cats and dogs into different breeds. The model uses transfer learning with a pre-trained ResNet50 architecture. The objective is to accurately identify the breed of the animal in the given image.

Dataset

The dataset comprises images of various cat and dog breeds. The images are stored in the ./images directory, and the file names follow the format <breed_name>_<index>.jpg. The following breeds are included:

• Cat Breeds:

  • Abyssinian
  • Bengal
  • Berman
  • Bombay
  • British Shorthair
  • Egyptian Mau

• Dog Breeds:

  • American Bulldog
  • American Pit Bull Terrier
  • Basset Hound
  • Beagle
  • Boxer
  • Chihuahua
  • English Cocker Spaniel
  • English Setter
  • German Shorthaired
  • Great Pyrenees

Preprocessing

1. Image Loading: Images are loaded using tensorflow.keras.preprocessing.load_img, and each image is resized to 224x224 pixels using TensorFlow's resize_with_pad method.

2. Label Encoding: The breed names are extracted from the file names, and each breed is encoded into a numerical label (e.g., Abyssinian = 0, Bengal = 1, etc.).

3. One-Hot Encoding: The labels are one-hot encoded to be used in the categorical classification.

4. Data Splitting:

   • The dataset is split into training, validation, and test sets using an 80-20 split for training and testing, and then 25% of the training data is used for validation.

Model Architecture

The model is built using the ResNet50 architecture with the following components:

• Data Augmentation: A sequential model is used to apply random flips and rotations to the input images.
• Transfer Learning:
  • A pre-trained ResNet50 model (trained on ImageNet) is used without the top layer (include_top=False) and with pooling='avg'.
  • The ResNet50 model's layers are frozen to prevent them from being trained further.
• Prediction Layer: A dense layer with 16 neurons and a softmax activation function is added for classifying the input image into one of the 16 classes (breeds).

Model Compilation

The model is compiled with the following configurations:

• Optimizer: Adam optimizer.
• Loss Function: Categorical cross-entropy.
• Metrics: Accuracy.

Training

The model is trained on the training set for 10 epochs, and the performance is evaluated on the validation set.

Evaluation

After training, the model is evaluated on the test set to determine its accuracy. Predictions are also made on the test set, and the results can be further analyzed.

Usage

To run the project, follow these steps:

1. Install Dependencies: Ensure you have Python installed along with the necessary libraries.

   pip install numpy pandas matplotlib tensorflow

2. Run the Script: Execute the script in a Python environment to train the model and evaluate its performance.

3. Visualize the Results: The training and validation accuracy/loss can be visualized using matplotlib to understand the model's learning curve.

Conclusion

This project demonstrates the use of transfer learning with ResNet50 for image classification tasks. The performance of the model can be improved with more data, better preprocessing, or fine-tuning the ResNet50 layers.

---

Author: Sahil Awatramani

---