Cat Image Classification: Project Overview

This project builds a deep learning model to automatically classify various cat (Felis) species from their images.

• Utilizing the Felis Taxonomy Image Classification dataset from Kaggle.
• Performing data preprocessing and image augmentation to enhance model generalization.
• Employing transfer learning technique on a pre-trained model for improved accuracy.
• Developing a user-friendly Flask API for real-time species prediction based on image inputs.

Code and Resources Used

Python Version: 3.12
Packages: numpy, pandas, matplotlib, tensorflow, tensorflow_datasets, flask, pillow
Flask API Setup:

• pip install -r requirements.txt
• conda env create -n <ENVNAME> -f environment.yaml (Anaconda Environment)

Dataset: https://www.kaggle.com/datasets/datahmifitb/felis-taxonomy-image-classification/data

Getting Data

The project utilizes the Felis Taxonomy Image Classification dataset, containing 519 JPG images of seven different cat species:

• Domestic cat (F. catus)
• European wildcat (F. silvestris)
• Jungle cat (F. chaus)
• African wildcat (F. lybica)
• Black-footed cat (F. nigripes)
• Sand cat (F. margarita)
• Chinese mountain cat (F. bieti)

Data Preprocessing

• Data Split: The dataset was divided into 80% for training and 20% for testing to ensure the model learns well and is evaluated fairly.
• Image Preprocessing: All images were resized to 224x224 pixels so they are perfectly uniform for the computer to read.
• Image Data Augmentation: To make the model more adaptable and prevent overfitting (getting too memorized on the training images), these techniques were applied:
  • Image Rotation
  • Image Translation (shifting)
  • Image Flipping
  • Contrast Adjustment

Model Building

The brain of this project relies on EfficientNetB0, a highly powerful pre-trained image recognition model. We adapted it using transfer learning (taking a model that already knows how to see shapes and colors and teaching it specifically about cats).

Here is the blueprint of our network:

Model Evaluation

The model's performance is measured using categorical cross-entropy (a score of how confident it is with its guesses) and optimized using the ADAM algorithm (the math tool that helps the model learn from its mistakes).

The training results are shown below:

Productionization

To bring the model to life, a user-friendly API was developed using Flask. This allows the model to act as a backend service: you upload a cat photo, and it instantly sends back the predicted species name.