FFD: Figure and Formula Detection from Document Images

Junaid Younas, Syed Tahseen Raza Rizvi, Muhammad Imran Malik, Faisal Shafait, Paul Lukowicz, Sheraz Ahmed

In: IEEE (Hrsg.). 2019 International Conference on Digital Image Computing: Techniques and Applications (DICTA). International Conference on Digital Image Computing Techniques and Applications (DICTA-2019) 21st December 2-4 Perth WA Australia IEEE 2019.


In this work, we present a novel and generic approach (FFD) to detect the formulae and figures from document images. Our proposed method employ traditional computer vision approaches to aid the deep-learning models in order to improve the performance in comparison to their conventional counterparts. We transform input images by applying connected component analysis (CC), distance transform and color transform and stacked them together to generate input image for deep models. Best results produced by FFD for figure and formula detection are with f1-score of 0.906 and 0.905, respectively. We also propose a new dataset for figures and formulae detection to further push the boundaries for research community in this direction.

FFD.pdf (pdf, 4 MB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence