Unsupervised learning predicts human perception and misperception of gloss.

"author contributions conceptualization k r s r w f and b l a; methodology k r s and r w f; software k r s; formal analysis k r s; investigation k r s; resources r w f; writing--original draft k r s and r w f; writing--review and editing k r s r w f and b l a; visualization k r s; supervision r w f; funding acquisition r w f and k r s data availability all human and model data are available on zenodo at 10 5281/zenodo 4495586.; code availability the custom analysis code that supports the findings of this study is available on zenodo at 10 5281/zenodo 4495586. data availability all human and model data are available on zenodo"

"we used the implementation from ref of the pixelvae architecture available at https://github com/ermongroup/generalized-pixelvae ."

"Competing interests The authors declare no competing interests."

"This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; project number 222641018–SFB/TRR 135 TP C1), the Australian Research Council, the Hessisches Ministerium für Wissenschaft und Kunst (HMWK; project ‘The Adaptive Mind’), by an award from the European Research Council (ERC; Consolidator Award ‘SHAPE’—project number ERC-CoG-2015-682859) to R.W.F. and by an Alexander von Humboldt Postdoctoral Research Fellowship to K.R.S. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. We thank A. Schmid and K. Doerschner for sharing code to implement the highlight feature measurement model and thank K. Gegenfurtner and J. Todd for comments on earlier versions of this manuscript."