Cat vs Dog Classifier using SVM | Prodigy Internship Task 3

Cat vs Dog Image Classifier using SVM

Internship: Machine Learning Virtual Internship by Prodigy Infotech

Task: PRODIGY_ML_03

Dataset : https://www.kaggle.com/competitions/dogs-vs-cats

P.S : I did nt upload dataset and pickle file as it's size breaks github upload limit.

📌 Project Description

This project implements a binary image classifier that distinguishes between cats and dogs using the Support Vector Machine (SVM) algorithm.

It is the third task of the internship and provided valuable hands-on experience in machine learning workflows and SVM model tuning.

🧠 Steps Followed

1. Labeled every image in the train folder based on file name prefixes.
2. Converted images to grayscale for uniform color structure.
3. Extracted HOG (Histogram of Oriented Gradients) features for robust representation.
4. Split data into 80% training and 20% testing, with shuffling.
5. Created a smaller subset of the data to run GridSearchCV for hyperparameter tuning.
6. Trained the final model using the best parameters from GridSearchCV.
7. Achieved an accuracy of ~78% on the test set.
8. Displayed and predicted test images directly in the Jupyter Notebook.
9. Saved the trained model using pickle for later use.

🖼️ GUI Feature

Used Tkinter to build a graphical user interface allowing users to:

• Upload an image
• Predict whether the image is of a cat or a dog

📚 Learnings

This project helped me understand key machine learning concepts such as:

• Feature extraction (HOG)
• Train-test split
• Model selection and evaluation
• Hyperparameter tuning using GridSearchCV
• Model deployment via GUI

If you have come this far, checkout my other projects also, if you want you can connect me.

Author: Mithun