PANOPTES: Frozen Study Specification and Analysis Plan for Budgeted PPE Inspection
This deposit records the study specification, evaluation protocol, development-informed analysis plan and file identities for PANOPTES (Policy-Aware Next-Observation Planning and Tool-based Evidence Selection for Personal Protective Equipment Monitoring). The study investigates whether selective inspection improves PPE detection under a bounded inference budget, and how improvements depend on action selection, evidence fusion and the evaluation metric. The registered pooled study includes five detector configurations, three training seeds and separate training, validation, calibration and final-test partitions. The reserved final-test partition contains 680 images. The archive contains the study specification, protocol, analysis clarification and SHA-256 identities of the specification and frozen partitions. Dataset images, annotations, model weights and credentials are excluded. Development experiments and training preceded this deposit. The analysis clarification was written after development audits. This record establishes a public timestamp for the deposited documents; it does not constitute prospective preregistration of the earlier experiments. At preparation of this deposit, the reserved final-test partition remained unscored. Selected checkpoints, calibrated thresholds, planner prompts and final execution settings require a subsequent evaluation seal. Author and sole contributor: James Afful, Center for Computational Research and Education (CCRE). No external funding supported this work. The author declares no competing interests.
Authors
- James Afful
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-15
- DOI
- https://doi.org/10.5281/zenodo.24247354
- Primary Topic
- Advanced Neural Network Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00