MapTD | EAST-inspired Tensorflow-based Text Detector | Machine Learning library

MapTD is a C++ library typically used in Artificial Intelligence, Machine Learning, Deep Learning, Pytorch, Tensorflow applications. MapTD has no bugs, it has no vulnerabilities, it has a Strong Copyleft License and it has low support. You can download it from GitHub.

MapTD is a convolutional neural network model for text detection written in Tensorflow. It is engineered for operating on mixed text/graphics documents, such as maps. The code is a derivative of @argman's implementation of EAST: An Efficient and Accurate Scene Text Detector by Zhou et al. Like argman's EAST code, the network backbone is Resnet-50, with a dice loss (whereas the EAST paper uses PVANet with balanced cross-entropy loss). This model uses a semantically-grounded orientation for the rectangles, so that it learns to produce rectangle sides in a specific order (i.e., text baseline, right, text top, left).