TechMB: 947 VQA Pairs for Manufacturing Evaluation from Technical Drawings

947 question-answer pairs form a Visual Question Answering benchmark for evaluating the manufacturability of objects based on 180 distinct technical drawings. The dataset was created by author WSKL and was last updated on 2025-10-14. It is designed as a domain-specific test for Vision Language Models.

Use Cases

• Benchmarking Vision Language Models on domain-specific Visual Question Answering tasks based on technical drawings.
• Training models to understand and evaluate manufacturability constraints from engineering diagrams.
• Developing AI assistants for design review by interpreting questions about technical drawings.

Strengths

• Contains 947 specific question-answer pairs for structured evaluation.
• Based on 180 distinct technical drawings, providing a foundation for visual reasoning.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

The technical drawings were developed from a selection of objects, as referenced in the description.

Freshness

Last updated 2025-10-14 07:50:32; freshness should be verified.