גהאד אלצאנע

This technical report presents a Fully Convolutional Network based method for layout analysis of benchmarking dataset provided by the competition. The document image is segmented into text and non-text zones by dense pixel prediction. Convolutional part of the network can learn useful features from the document images and is robust to uncontrained layouts. We have evaluated the zone segmentation with average black pixel rate, over-segmentation error, under-segmentation error, correct-segmentation, missed-segmentation error, false alarm error, overall block error rate whereas the zone classification with precision, recall, F1-measure and average class accuracy on both pixel and block levels.