Explainable Computer Vision with Grad-CAM

Building powerful Computer Vision-based apps without deep expertise has become possible for more people due to easily accessible tools like Python, Colab, Keras, PyTorch, and Tensorflow. But why does a computer classify an image the way that it does? This is a question that is critical when it comes to AI applied to diagnostics, driving, or any other form of critical decision making. In this episode, I'd like to raise awareness around one technique in particular that I found called "Grad-Cam" or Gradient Class Activation Mappings. It allows you to generate a heatmap that helps detail what your model thinks the most relevant features in an image are that cause it to make its predictions. I'll be explaining the math behind it and demoing a code sample by fairyonice to help you understand it. I hope that after this video, you'll be able to implement it in your own project. Enjoy! TWITTER: https://bit.ly/2OHYLbB WEBSITE: https://bit.ly/2OoVPQF INSTAGRAM: https://bit.ly/312pLUb FACEBOOK: https://bit.ly/2OqOhx1 Subscribe for more educational videos! It would mean a lot to me. One or more of the many slides i used to help convey concepts to people didn't include the name of the original source, the "Fraunhofer institute". This was not intentional, i literally forgot. I am one guy, and i'm making a sincere effort to invite more people into the AI community through educational videos while simultaneously crediting all the giant shoulders that my explanations stand on. I'll do better next week! Here are the 4 explainable AI videos on Youtube i liked the most: https://www.youtube.com/watch?v=B-c8tIgchu0 https://www.youtube.com/watch?v=CY3t11vuuOM https://www.youtube.com/watch?v=4pgLsDzrlB8 https://www.youtube.com/watch?v=I0yrJz8uc5Q Grad-CAM Demo: http://gradcam.cloudcv.org/ FairyOnIce's Grad-CAM code: https://github.com/FairyOnIce/FairyOnIce.github.io This implementation looks pretty cool as well: https://github.com/jacobgil/keras-grad-cam Are you a total beginner to