Using Pretrained Model for Wolf vs Dog Classification

In this project pretrained model googlenet is used to classify between dog and wolf images

Project Title:

Image Classification Using Transfer Learning with GoogLeNet

Project Summary:

This project fine-tunes the GoogLeNet pretrained model for image classification on a custom dataset using transfer learning.

1. Data Preparation:

Images were loaded using `imageDatastore`, organized by folder labels, and split into 80% training and 20% validation sets.

2. Network Customization:

GoogLeNet was loaded, and its final fully connected and classification layers were replaced to match the number of classes in the custom dataset. Learning rates were adjusted for faster training on the new layers.

3. Data Augmentation:

An augmented image datastore resized images to 224x224 and applied random reflections and translations to prevent overfitting.

4. Training:

The network was trained with a small learning rate and validated after each epoch.

5. Evaluation:

The model classified validation images, and confusion matrices (both raw and normalized) were plotted to assess accuracy.

This project successfully applied transfer learning for image classification using GoogLeNet.