FCR: Flow Chart Recognition Network for Program Synthesis

Abstract

Program synthesis is the task to automatically generate programs based on user specification. In this paper, we present a framework that synthesizes programs from images of flow chart that serve as accurate and intuitive specifications. In order doing so, we propose a deep neural network called FCR that recognizes graph structure from its image. FCR is trained end-to-end, which can predict edges and nodes of the flow chart simultaneously. Experiments show that the accuracy to synthesize a program is 66.4%, and the accuracy to recognize edges and nodes are 94.1% and 67.9%, respectively. On average, it takes about 60 milliseconds to synthesize a program.

Tasks

Reproductions