DeepDP is a multi-level convolutional neural network based on the U-Net architecture. Its primary goal is to process binary (black and white) images of hand-drawn diagrams and output a map that identifies each pixel as: : Written words or labels.

: Graphical elements like boxes, circles, and connectors. Background : Empty space on the page.

Deep handwritten diagram segmentation - IET Journals - Wiley

Try Impetus DEFENSE free for three months with full access to all features, support, and training. Our team will guide you through setup, onboarding, and evaluation so you can focus on testing the software for your needs.