This project aims to develop a deep learning model for dog breed classification using the ResNet50 architecture. The ResNet50 is a pre-trained convolutional neural network (CNN) model that is commonly used in image classification tasks.
The dataset can be downloaded from the Kaggle. The dataset consists of 5 common dog's breeds:
- French Bulldog (208 images)
- German Shephard (247 images)
- Golden Retriever (213 images)
- Poodle (182 images)
- Yorkshire Terrier (180 images)
We used transfer learning to fine-tune the ResNet50 model on the dataset. Transfer learning is a technique in deep learning where a pre-trained model is used as a starting point for a new task. By using a pre-trained model, we can reduce the training time and improve the accuracy of the model.
We first initialized the ResNet50 model with pre-trained weights on the ImageNet dataset. We then removed the last layer of the model and replaced it with a new fully connected layer with 120 nodes, corresponding to the number of dog breeds in the dataset. We froze the weights of the pre-trained layers and only trained the new fully connected layer. We used the categorical cross-entropy loss function and the Adam optimizer to train the model.
After training the model on the training set and evaluating it on the validation set, we achieved an accuracy of 98.29%. We then evaluated the model on the testing set and achieved an accuracy of 98.44%.
In this project, we developed a deep learning model for dog breed classification using the ResNet50 architecture. We achieved an accuracy of 98.44% on the testing set. This model can be used for various applications, such as identifying dog breeds in images and assisting in the development of pet-related apps.