Using computer vision to measure design similarity: An application to design rights

Abstract

Competition among firms has increasingly been through design. We show how computer vision algorithms can be leveraged to measure the visual similarity of design rights across large data sets of product design images. In particular: we extract and standardize 716,168 unique design images included in US design patents (1976–2023); adapt the structural similarity index measure to quantify design similarities between images; and rigorously validate the resulting measure of design rights similarity. We then use that measure to produce novel empirical evidence that a design space's similarity density exhibits an inverted U-shape with respect to the likelihood of that space's design rights being litigated—a relationship proposed previously but never tested. Our design rights similarity measure should facilitate the exploration of new research questions in the fields of design rights, innovation, and strategy. We grant open access to our code and data resources to encourage research in such fields. • We develop a method for measuring design similarity via SSIM on 716,168 US design patent images (1976–2023). • We validated the similarity measure through three strategies: (1) correlational validity (2) predictive validity; and (3) human validity. • We conceptualize design space similarity density (DSSD) as the degree of visual similarityin a design space and show its role in shaping firms’ litigation strategy. • We show an inverted U-shaped relationship between DSSD and litigation. • We release the full dataset and code to support future research on design rights, innovation and IP strategy using visual similarity measures.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 5 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.