Viet Handwriting Gemini VQA: Vietnamese Handwriting Images with Annotations

Over 8,700 labels and descriptions were generated for 1,252 Vietnamese handwriting images using the Gemini 1.5 Flash model. The dataset was created by 5CD-AI from the train splits of the Cinnamon AI Challenge and UIT-HWDB datasets. It was last updated on Hugging Face in August 2024.

Use Cases

• Train Vietnamese handwriting recognition models based on the annotated images.
• Develop visual question answering systems for handwritten text using the generated Q&A pairs.
• Benchmark multimodal AI models on a Vietnamese-language task.
• Fine-tune large language models for understanding handwritten documents.

Strengths

• Contains 1,252 Vietnamese handwriting images.
• Includes over 8,700 generated labels, descriptions, and Q&A pairs.
• Leverages a specific advanced model, Gemini 1.5 Flash, for annotation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the generated annotations is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the source datasets and the generative model used for annotation.

Provenance

Source

Cinnamon AI Challenge - Handwriting address and UIT-HWDB datasets.

Collection Method

Images were analyzed and annotated using Gemini 1.5 Flash model for Visual Question Answering.

Freshness

Last updated 2024-08-25 15:58:31; freshness should be verified.

Geography

Vietnam