Open Images V5 Text Annotation and Yet Another Mask Text Spotter
This work provides a large-scale dataset and a model for text spotting, which is useful for computer vision researchers and practitioners, but it is incremental as it builds on existing methods like Mask-RCNN.
The authors introduced text annotations for the Open Images V5 dataset, which they claim is the largest publicly available manually created text annotation, and trained a Mask-RCNN-based model called YAMTS that achieves competitive or superior performance on standard text spotting datasets like ICDAR2013, ICDAR2015, and Total-Text.
A large scale human-labeled dataset plays an important role in creating high quality deep learning models. In this paper we present text annotation for Open Images V5 dataset. To our knowledge it is the largest among publicly available manually created text annotations. Having this annotation we trained a simple Mask-RCNN-based network, referred as Yet Another Mask Text Spotter (YAMTS), which achieves competitive performance or even outperforms current state-of-the-art approaches in some cases on ICDAR2013, ICDAR2015 and Total-Text datasets. Code for text spotting model available online at: https://github.com/openvinotoolkit/training_extensions. The model can be exported to OpenVINO-format and run on Intel CPUs.