Digitize-PID: 500 Synthetic P&ID Images for Symbol Detection

Digitize-PID is an annotated synthetic dataset of 500 Piping and Instrumentation Diagrams (P&IDs) containing only symbols for object detection tasks. The dataset incorporates different types of noise and complex symbols. It was created by Paliwal, S., Jain, A., Sharma, M., & Vig, L. (2021) and is hosted on Hugging Face by user hamzas.

Use Cases

• Train object detection models based on the annotated P&ID symbols.
• Benchmark model robustness based on the synthetic noise and complex symbols mentioned in the description.
• Develop digitization pipelines for technical drawings based on the P&ID symbol annotations.

Strengths

• Contains 500 annotated images.
• Specifically designed for object detection tasks on symbols.
• Incorporates varied noise and complex symbols for robustness testing.

Limitations

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

Provenance

Source

Original data repository referenced but not named; dataset is hosted on Hugging Face.

Collection Method

Synthetic dataset generation.

Time Range

Associated with a 2021 research paper.

Freshness

Last updated 2025-08-06 13:36:01; freshness should be verified.

Geography

Spatial coverage is unknown.