How to Classify a Design Using Convolutional Neural Network Models

With convolutional neural networks, there are methods where you can actually query the model to tell you exactly what parts of an image were the most important in making its characterization. We used a technique called a gradcam that creates heap maps, which show which sections of the image the model considered the most important and which sections were less important. Using this approach, you can see, you know, it gives you a direct feedback as to we know, what sections, or what design elements a particular sirt are the critical elements in picking out that type. And just to show it graphically, ris going to expand a little bit on what chris said.