Clothing Size Recommendations from Sparse Body Measurements and Photographed Charts
Year: 2026
Type: preprint
Abstract: Choosing a garment by nominal size is unreliable across retailers, garment types and markets. This source-level study traces the original Android Size Guide from sparse body inputs (height, weight, gender, region and optional direct measurements) through deterministic body-dimension estimation, chart selection, validation and fit scoring. It also examines the optional on-device Google ML Kit OCR route that reconstructs size-table cells from a photograph before recommendation. The manuscript identifies implementation limits, including the bypassed trained predictor, inconsistent regional assumptions and unmeasured OCR and fit accuracy. It proposes an empirical evaluation protocol but reports no invented benchmark results. The accompanying ZIP contains a selected, byte-identical snapshot of 62 original Kotlin files and a SHA-256 manifest from the Vanexa Android application; it is not a standalone build. The PDF and LaTeX manuscript are licensed CC BY 4.0; the companion code snapshot is MIT licensed. Source repository: https://github.com/ikbal-causyne/clothing-size-recommendations.
Author: Fadilah, Ikbal
Language: English
Topic: Mobile and Web Applications
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7223677823