Structural evidence learning for fine grained classroom behavior detection with coherence guided confidence refinement

Abstract Fine-grained classroom behavior detection is challenging because visually similar behaviors rely on weak and distributed visual cues. We propose PCSF-ONE-YOLO, which reorganizes candidate features into latent structural nodes through Phase-Coupled Structure Folding (PCSF) and uses structural coherence for confidence refinement. The method requires no additional structural annotations and preserves the native localization formulation. Experiments on SCB5 and SCBehavior with dataset-specific training show competitive detection performance and modest mean gains over YOLOv9e on selected metrics. On SCB5, the full model has lower mean confidence errors than YOLOv9e on shared true-positive detections, while standard detection-level calibration does not improve at a confidence threshold of 0.25. The three-seed analysis does not establish a statistically significant AP improvement. The results support structural evidence modeling as a complementary mechanism for fine-grained classroom behavior detection.

Authors

Institutions

Publication Details

Journal
Discover Computer Vision
Published
2026-11-13
DOI
https://doi.org/10.1007/s44584-026-00011-y
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Structural evidence learning for fine grained classroom behavior detection with coherence guided confidence refinement

Guoming Chen, Xiongyong Zhu, Jianchun Xu, Ruqi Zhou et al.
5 links
Discover Computer Vision
Advanced Neural Network Applications
article

Structural evidence learning for fine grained classroom behavior detection with coherence guided confidence refinement

Guoming Chen, Xiongyong Zhu, Jianchun Xu, Ruqi Zhou, Shijin Lin
article en

Abstract

Abstract Fine-grained classroom behavior detection is challenging because visually similar behaviors rely on weak and distributed visual cues. We propose PCSF-ONE-YOLO, which reorganizes candidate features into latent structural nodes through Phase-Coupled Structure Folding (PCSF) and uses structural coherence for confidence refinement. The method requires no additional structural annotations and preserves the native localization formulation. Experiments on SCB5 and SCBehavior with dataset-specific training show competitive detection performance and modest mean gains over YOLOv9e on selected metrics. On SCB5, the full model has lower mean confidence errors than YOLOv9e on shared true-positive detections, while standard detection-level calibration does not improve at a confidence threshold of 0.25. The three-seed analysis does not establish a statistically significant AP improvement. The results support structural evidence modeling as a complementary mechanism for fine-grained classroom behavior detection.

Discover Computer VisionVol. 1(1)
Guangdong University of Education (CN)
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon