Canvas Method

What is the Canvas Method for Object Detection Models? The Canvas Method is a technique used to conduct inference attacks on object detection models. It is a way to draw predicted bounding box distributions on an empty canvas, which is initially set to an image of 300$\times$300 pixels in size. How does the Canvas Method work? The process begins with an attack model input, which is used to create a canvas with every pixel having a value of zero. The predicted boxes are drawn on the canvas us