Egyptian Hieroglyphic Signs Segmentation with Orientation

Egyptian Hieroglyphic Signs Segmentation with Orientation Datasets for Ancient Egyptian Hieroglyphic Research Egyptian Hieroglyphic Signs Segmentation with Orientation (SS) Dataset Overview: The Signs Segmentation (SS) Dataset comprises 300 images, each containing a single line of ordered Ancient Egyptian hieroglyphic signs. These images were automatically cropped from segmented lines within the HLA Dataset using our trained layout analysis models. The SS Dataset is specifically designed for the task of segmenting individual hieroglyphic signs within a line of text. Data Generation and Annotation: The lines of hieroglyphs in the HLA Dataset were processed using our trained models to automatically extract individual lines. These cropped line images form the basis of the SS Dataset. Each image in this dataset was then manually annotated using the CVAT platform with polygonal segmentation masks for three distinct classes: “Left Sign”, “Right Sign”, and “Dual Sign” (representing ligatures or signs that visually merge) that present the orientation of signs. Key Statistics: Total Images: 300 Content: Single lines of ordered hieroglyphic signs Annotation Classes: 3 (“Left Sign”, “Right Sign”, “Dual Sign”) Annotation Type: Polygon Segmentation Masks Potential Uses: This dataset is well-suited for training and evaluating models for: Individual hieroglyphic sign segmentation within a line of text. Distinguishing between individual and joined signs. Classify the orientation of signs. Json annotation files "in coco format": Train: 272 images. Validation: 28 images. Test: 0 images.