This repository contains the code and resources for an image classification project which is part of the AWS ML & AI Programming with Python Nanodegree Scholarship Program. The goal of this project is to use a pre-trained image classifier to classify images into different categories, including dogs, breeds of dogs, and non-dogs.
The image classification project utilized various CNN (Convolutional Neural Network) models, namely VGG, AlexNet, and ResNet, to classify images. The code implemented a pipeline that loaded the images, processed them, and passed them through the trained models to generate predictions. The predictions were then evaluated to determine the accuracy of the models.
- Implemented image classification using VGG, AlexNet, and ResNet models
- Loaded and preprocessed images for classification
- Evaluated the accuracy of the models on a given dataset
- Printed summary statistics and results
The classification results for the three models are as follows:
VGG Model Architecture: Achieved 100% accuracy in classifying dogs, breeds, and non-dogs. Total Elapsed Runtime: 8 seconds.
AlexNet Model Architecture: Achieved 100% accuracy in classifying dogs, breeds, and non-dogs. Total Elapsed Runtime: 1 second.
ResNet Model Architecture: Achieved 100% accuracy in classifying dogs, breeds, and non-dogs. Total Elapsed Runtime: 4 seconds.
In conclusion, all three models performed exceptionally well in classifying the given images with perfect accuracy. However, considering both accuracy and efficiency, the AlexNet model architecture stood out as the most efficient, with the shortest runtime of only 1 second. Therefore, the AlexNet model is recommended for classification tasks in scenarios where runtime is a crucial factor.
This project is licensed under the MIT License.
This image classification project is based on the foundations provided by the Udacity AI Programming with Python Nanodegree program.
Special thanks to the Udacity instructors and mentors for their guidance and support throughout the project.