Handwritten Circuit Diagrams with Bounding Box and Polygon Annotations

The CGHD dataset provides images of hand-drawn electrical circuit diagrams with ground-truth annotations for object detection, instance segmentation, and text detection. Created by lowercaseonly, it is intended to facilitate the automated extraction of electrical graphs from handwritten sources. The dataset was last updated on February 27, 2026.

Use Cases

• Train an object detection model to locate circuit components using the provided bounding box annotations.
• Perform instance segmentation on hand-drawn diagrams using the polygon annotation files.
• Develop a text detection system for handwritten labels within the circuit diagram images.
• Evaluate image processing algorithms for extracting electrical graphs from the annotated diagram images.

Strengths

• Dataset is specifically designed for multiple computer vision tasks: object detection, instance segmentation, and text detection.
• Provides multiple annotation types (bounding boxes and polygons) for the same set of hand-drawn diagram images.
• Last update recorded as February 27, 2026, indicating recent maintenance.

Limitations

• The total number of images and annotation files is unknown, making it difficult to assess scale.
• Specific file formats for images and annotations are not provided in the input.
• Potential for label noise or inconsistency inherent in hand-drawn diagrams is not addressed.

Provenance

Source

lowercaseonly on Hugging Face

Collection Method

Collection of hand-drawn electrical circuit diagrams with manually created annotations.

Freshness

Last updated on 2026-02-27.