Explainable AI: Scene Classification and GradCam Visualization

In this 2-hour guided project, you will delve into the world of Explainable AI and scene classification using deep learning techniques. The project will equip you with the knowledge and practical skills to train a deep learning model to predict the type of scenery in images. Additionally, you will gain insights into the application of Grad-Cam, a technique used to elucidate AI model decision-making processes. The content of the project is structured to provide a comprehensive understanding of the underlying theories and practical implementation of Convolutional Neural Networks (CNNs) and Residual Nets. Through hands-on exercises, you will build a deep learning model based on CNN and Residual blocks using Keras with Tensorflow 2.0 as a backend, enabling you to visualize activation maps used by CNNs for predictions and deploy the trained model using Tensorflow Serving.

Key Learning Objectives:

• Understanding the theory and intuition behind Deep Neural Networks, Residual Nets, and Convolutional Neural Networks (CNNs)
• Building a deep learning model based on CNN and Residual blocks using Keras with Tensorflow 2.0 as a backend
• Visualizing the Activation Maps used by CNN to make predictions using Grad-CAM
• Deploying the trained model using Tensorflow Serving