Transfer Learning

 

As part of a graded assignment for the 4th course in the Deep Learning Specialization offered by Stanford online and deeplearning.ai, I implemented transfer learning on a pre-trained Convolutional Neural Network (CNN) to build an Alpaca/Not Alpaca classifier.

​

The CNN, MobileNetV2, was designed to provide fast and computationally efficient performance. It's been pre-trained on ImageNet, a dataset containing over 14 million images and 1000 classes.

​

For transfer learning, I created a dataset of alpacas from a directory provided by the course assistants. Next, I preprocessed and augmented the data using TensorFlow's Sequential API, applying various image transformations like rotations and flipping to enhance the dataset's diversity and robustness. Then, I adapted the pre-trained model to this new data, using the Functional API along with MobileNet for efficient feature extraction. Finally, I fine-tuned the classifier's final layers, adjusting parameters and layers to improve the model's accuracy and performance.