Large Language Model and Synthetic Dataset Based Occupant Clothing Insulation Recognition
Bibliographic record
Abstract
With individuals spending over 85% of their time indoors, maintaining Thermal Comfort (TC) is vital for well-being and productivity.The Predicted Mean Vote (PMV), a key metric for assessing thermal sensation, integrates four environmental parameters and two personal factors: Clothing Insulation (CI) and Metabolic Rate (MR).While environmental factors are measurable through sensors or simulations, personal factors are often oversimplified, reducing the accuracy of TC assessments.Recent advancements in Machine Learning (ML) and sensor technologies offer innovative methods for estimating these personal factors.However, current approaches rely on datasets collected labor-intensively and privacy-invasively, limiting their scalability and quality.Moreover, despite recent impressive achievements across many domains of Large Language Models (LLMs), the potential of zero-shot LLMs for personal factor recognition remains unexplored.To address these limitations, this study introduces a high-quality synthetic dataset developed using Unreal Engine (UE).The dataset includes 2,250 monocular RGB videos paired with expert-extracted keyframe images, featuring single occupant engaged in diverse activities.Each videoimage pair is meticulously annotated with clothing types and corresponding overall CI values, adhering to ASHRAE standards.Leveraging this dataset, a state-of-the-art (SOTA) LLM is employed for zero-shot CI recognition, achieving an overall success rate of 75.6% in clothing recognition and a Mean Absolute Error (MAE) of 0.03 in CI predictions.These results validate the dataset's quality and demonstrate the potential of LLMs in capturing occupant-specific personal factors.This research highlights a pathway for more precise and scalable TC evaluations, paving the way for smarter and more occupant-centric indoor environments.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".