Reproducibility Package for "Visual-Taxonomic Informativeness: A Knowledge-Based Framework for Assessing Recoverable Evidence in Citizen-Science Image Repositories"

This record provides the reproducibility materials accompanying the manuscript “Visual-Taxonomic Informativeness: A Knowledge-Based Framework for Assessing Recoverable Evidence in Citizen-Science Image Repositories.”The package supports the construction, estimation, prediction, and downstream evaluation of Visual-Taxonomic Informativeness (VTI), an image-level measure of recoverable fine-grained visual evidence informed by expert-defined visual ambiguity among taxa.The deposited materials include source code, experimental manifests, data registries, expert-informed visual-ambiguity definitions, selected ONNX backbone models, trained analysis artifacts, derived image embeddings, prediction files, result tables, statistical summaries, and generated figures supporting the manuscript and its supplementary material.The analysis covers a balanced collection of 120 fungal species and 120,000 citizen-science images. The original images are not redistributed because they remain subject to the licensing and availability conditions of the source platform. The provided manifests and registry information identify the data used in the study and support reconstruction from the original source where permitted.Third-party model components remain subject to their respective licenses. Licensing and attribution information is included in the deposited package.