Wordart: Artistic Text Recognition for CornerTransformer (ECCV 2022)

Wordart is a dataset for artistic text recognition developed by xdxie for the CornerTransformer model presented at ECCV 2022. It focuses on non-standard text layouts and stylized fonts, providing a benchmark for OCR systems beyond standard document or scene text.

Use Cases

• Training CornerTransformer models for artistic text recognition
• Benchmarking OCR performance on stylized and non-linear fonts
• Implementing contrastive learning for text-image alignment

Strengths

• Associated with ECCV 2022 Oral publication
• Targets artistic text recognition, a niche OCR domain
• Licensed under Apache-2.0 for open use

Limitations

• Unknown record count and total file size
• Lack of detailed column or metadata documentation in the source repository

Provenance

Source

xdxie (GitHub) and ECCV 2022 publication

Freshness

Last updated in March 2023 following the ECCV 2022 publication.