Implementation of EAST: An Efficient and Accurate Scene Text Detector in MXNet.
The pipeline is simpler yet achieves almost the SOTA results. Among all nearly-SOTA models, EAST achieves F-score of 0.7820 at 13.2fps at 720p resolution. The current SOTA results (as of Oct 2017) in text detection is achieved by Fused Text Detection Network, which has a speed of 4 FPS on 848 ร 1500 images.