Synthtiger: Synthetic Scene Text OCR Dataset

Synthetic scene text images and text labels across multiple typographic and background categories. These samples are generated via the SynthTIGER engine to provide diverse typographic and environmental variations for optical character recognition.

Use Cases

• Train OCR models using the image and label columns
• Evaluate text recognition performance on synthetic distortions applied to the image data
• Fine-tune scene text detection algorithms using the generated image samples

Strengths

• Synthetic images paired with ground truth text labels
• Diverse font styles and background textures for scene text simulation
• Geometric transformations including perspective warping and blur