Dataset · jinlong666
Industry Document Specification Extraction Benchmark
jinlong666/IndustryBench-MIPU
A collection of Chinese technical tables and blueprint images for evaluating Multimodal Large Language Models (MLLMs).
What's inside
The dataset contains between 10k and 100k records featuring industrial product images, specification tables, nameplates, and technical blueprints in Chinese. The data evaluates how accurately multimodal models can extract structured features from disparate visual sources.
What you can do with it
- Evaluating MLLM performance on Visual Question Answering (VQA) tasks involving complex technical documents
- Testing OCR and attribute parsing systems for data extraction from blueprints and tables
- Tasks
- Image → text, Visual question answering
- Languages
- Chinese
- Size
- 10K–100K rows
- Downloads this month
- 21.4K
- Likes
- 1
How to load it
from datasets import load_dataset
ds = load_dataset("jinlong666/IndustryBench-MIPU")