Supporting Numerical Data for Fine-Grained Visual Recognition of Residential Air Conditioner Units

This repository contains the supporting numerical data and plotting scripts associated with the manuscript: Fine-Grained Visual Recognition of Residential Air Conditioner Units for Building Energy Management The deposited files contain the aggregate numerical results reported in the manuscript, including classification metrics, computational-efficiency measurements, and ablation-study comparisons. Repository contents The repository includes CSV files supporting the following experiments: Baseline architecture comparison Computational-efficiency comparison Activation-function comparison Classification-head comparison Learning-rate scheduler comparison Backbone architecture comparison Incremental ablation-study results Python scripts used to reproduce the corresponding plots are also included. Data provenance The CSV files contain the final aggregate numerical values reported in the manuscript tables and figures. The original per-image prediction outputs and complete per-epoch training logs were not retained. The deposited files therefore support reproduction of the reported tables and plots but do not constitute the complete original experimental records. Restricted image dataset The raw image dataset is not included in this repository. Field-acquired photographs were captured in private premises and may contain contextual details relating to the premises or their occupants. The contributors consented to the use of the photographs for research, model training, evaluation, and publication of the study results, but not to unrestricted public dissemination under an open licence. The dataset also contains images obtained from publicly accessible manufacturer and retail websites. These images are subject to third-party copyright and cannot be redistributed by the authors. Controlled access to eligible field-acquired images may be considered upon request, provided that the proposed use is compatible with the consent obtained and applicable data-protection requirements. File formats .csv: numerical data supporting manuscript tables and figures .py: Python scripts used to reproduce plots Software requirements The plotting scripts require Python 3 and the following packages: pandas matplotlib The Python scripts may be reused with attribution. The licence does not apply to the restricted raw image dataset or to third-party web-sourced images. Contact For questions regarding the supporting data, contact: Rafail Daskos Decision Support Systems Laboratory National Technical University of Athens Email: [e-mail]