cs-chan/Total-Text-Dataset

Total Text Dataset. It consists of 1555 images with more than 3 different text orientations: Horizontal, Multi-Oriented, and Curved, one of a kind.

What it solves

This project provides a word-level English curved text dataset designed to help machine learning models better detect and recognize text that is not straight, such as curved or arbitrary-shaped text in natural scenes.

How it works

It provides a comprehensive collection of images and corresponding ground-truth annotations. These annotations are available in multiple formats, including text-level, pixel-level, and .txt formats, to support various training and evaluation needs. The project also includes a baseline implementation and a guided annotation toolbox to assist in the scene text image annotation process.

Who it’s for

Researchers and developers working on scene text detection, arbitrary-shaped text reading, and computer vision tasks involving curved text recognition.

Highlights

  • Diverse Annotations: Offers ground-truth data in text, pixel, and text-file formats.
  • Evaluation Tools: Includes the DetEval evaluation protocol and scripts to benchmark detection performance.
  • Annotation Support: Provides a Guided Annotation toolbox for creating new scene text annotations.
  • Benchmarking: Maintains a detailed leaderboard tracking the performance of various state-of-the-art detection methods.

Related

  • Project
    PFCCLab/PPOCRLabel

    A semi-automatic graphic annotation tool for OCR that uses built-in PP-OCR models to automatically detect and recognize text for training dataset creation.

  • Project
    xiaofengShi/CHINESE-OCR

    A comprehensive pipeline for Chinese OCR in natural scenes, combining text orientation detection, CTPN for region detection, and CRNN for end-to-end character recognition.

  • Project
    inception-project/inception

    A machine-assisted text annotation platform for building annotated corpora with multi-layer schemes and knowledge-base entity linking.

  • Project
    vietanhdev/anylabeling

    An AI-powered image annotation tool that integrates YOLOv8 and Segment Anything to provide automatic labeling for computer vision datasets.

  • Project
    webis-de/small-text

    A Python library that provides ready‑to‑use active‑learning loops for text classification, supporting scikit‑learn, PyTorch, and Hugging Face transformers, with many query strategies, initialization methods, and stopping criteria.