Four benchmarks called AEC-Bench: where AI tests for construction fall short
Matthias Hornung, CTO of flinq, sorted the published benchmarks for AI in architecture, engineering and construction by year, language and data type. Two gaps stand out: apart from English, not a single European language is represented, and only a few benchmarks use 3D data as their main input. AI for construction is mostly tested on documents about buildings, rarely on the buildings themselves.
Read on LinkedIn