Multilingual Image Corpus
Svetla Koeva · Cognitive technologies · 2022
Abstract The ELG pilot project Multilingual Image Corpus (MIC 21) provides a large image dataset with annotated objects and multilingual descriptions in 25 languages. Our main contributions are: the provision of a large collection of highquality, copyright-free images; the formulation of an ontology of visual objects based on WordNet noun hierarchies; precise manual correction of automatic image segmentation and annotation of object classes; and association of objects and images with extended multilingual descriptions. The dataset is designed for image classification, object detection and semantic segmentation. It can be also used for multilingual image caption generation, image-to-text alignment and automatic question answering for images and videos.