Face Detection in Thermal Images with Improved Spatial Precision and Temporal Stability
Bibliographic record
Abstract
Thermal video can be used as a privacy-preserving and non-contact sensor for long-term health monitoring including respiratory activity. Face detection from thermal images is required to define the region of interest for automated respiration monitoring. In this study, we focus on thermal face detection using deep learning-based methods and transfer learning. First, YOLOv7, YOLOv7-tiny, and Detector Transformer (DeTr) object detection models were trained on an open thermal image dataset of faces. The weights from the pretrained models were transferred to a new model that was trained on our own target dataset. Results showed that transfer learning resulted in improved intersection-over-union (IoU) face detection performance. Moving beyond face detection in a single frame, we evaluated the stability of the trained face detection model with regard to the time-consistency of the detected bounding boxes in thermal videos. The DeTr model showed higher performance with 0.812 IoU and more stable predicted bounding boxes compared to YOLOv7 and YOLOv7-tiny. The proposed methods were also evaluated with regard to model size, as it pertains to viable deployment using edge computing, as part of a complete respiration rate estimation pipeline.
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 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| 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 | 0.001 |
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".