
We're building a labeled dataset from engineering diagrams. We need an annotator to label each diagram as a graph: every symbol marked with a classified bounding box, and every connection between symbols marked as an edge.
What you'll annotate (per sheet)
Node bounding boxes, each assigned one of 7 classes: valve, pump, instrumentation, general, tank, arrow, inlet/outlet. Routing/helper nodes: connector (line bends/junctions) and crossing (line crossings). Edges between nodes representing connectivity, each labeled solid or non-solid. Boxes drawn at full resolution, aligned to the supplied images. Deliverables
Annotations exported in [Label Studio JSON / CVAT XML / COCO] (final tool TBD — see below), in the same coordinate frame as the supplied images. We handle any downstream conversion; you provide clean, consistent labels matching our written guideline. Volume
~60 sheets total. We'll begin with a small paid pilot before the full batch. Required
Comfortable reading technical/engineering diagrams. Experience with an annotation tool that supports both bounding boxes and region-to-region relations (e.g., Label Studio or CVAT). Consistent, detail-oriented labeling; comfortable working on large images. Nice to have
Prior experience annotating for object-detection or graph/relationship ML datasets.
Software
Data Type
Task Types
Subject Matter / Industry
Share link