ankush-me/SynthText

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2146 stars · 624 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

SynthText is a Python tool for generating synthetic scene-text images with ground-truth bounding boxes, as described in the CVPR 2016 paper 'Synthetic Data for Text Localisation in Natural Images'. It renders text onto natural images using depth and segmentation maps, and a pre-generated dataset of ~800k synthetic images is also available.

Use cases

• generate synthetic training data for text detection models
• create scene text images with ground truth bounding boxes
• augment OCR training datasets with synthetic text
• render text onto natural background images
• train text localisation networks without manual annotation
• produce synthetic scene-text dataset like SynthText 800k

When to choose

• you need labeled scene-text training data for object detection or OCR models
• you want to reproduce the SynthText dataset or customize its generation
• you have background images with depth and segmentation maps available

When to avoid

• you need a ready-made OCR or text detection model rather than training data
• your background images lack depth maps and segmentation
• you need actively maintained software with modern Python packaging

Facets

library · maturity maintenance

image-processing data-generation computer-vision machine-learning computer-vision machine-learning image-processing deep-learning python windows synthetic-data scene-text text-detection ocr text-localisation dataset-generation cvpr-2016 linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="ankush-me/SynthText")