Total-Text

ImagesBSD-3Introduced 2017-01-01

Total-Text is a text detection dataset that consists of 1,555 images with a variety of text types including horizontal, multi-oriented, and curved text instances. The training split and testing split have 1,255 images and 300 images, respectively.

Source: Convolutional Character Networks Image Source: https://github.com/cs-chan/Total-Text-Dataset

Benchmarks

Scene Text Detection/F-Measure Scene Text Detection/Precision Scene Text Detection/Recall Scene Text Detection/FPS Text Spotting/F-measure (%) - No Lexicon Text Spotting/F-measure (%) - Full Lexicon