Original research visualizations

See how im2latex reads an equation

Yuntian Deng, Anssi Kanervisto, Jeffrey Ling, and Alexander M. Rush

Explore the network, attention alignments, and qualitative results from Image-to-Markup Generation with Coarse-to-Fine Attention (ICML 2017).

These are the original recorded model outputs. Select an example, then hover over its generated LaTeX tokens to inspect the attention over the input image.

Image-to-Markup encoder-decoder network architecture

Loading the original attention examples…

Loading the original qualitative results…

The alignment examples load a large data file and may take a moment. On a narrow screen, scroll within the visualization to see the full expression.