Cross‑device and cross‑modal arc detection: Transfer learning between EAST visible and WEST infrared diagnostics
Abstract
This work presents an engineering-oriented data processing and model training pipeline for EAST visible and WEST infrared data, enabling a systematic study of cross-device and cross-modal electric arc detection. A lightweight convolutional neural network is adopted as a cross-device arc detection model, and joint training and transfer learning experiments are conducted on EAST visible (VIS) and WEST infrared (IR) data. Single-modality baselines are first trained on EAST and WEST, then evaluated zero-shot on the opposite device, followed by a comparison of frozen-backbone and fine-tuned transfer for VIS→IR and IR→VIS.The results show that models trained from scratch on each diagnostic channel already achieve high precision and recall for arc detection, whereas direct zero-shot cross-modal transfer drives the arc recall on the opposite domain almost to zero. In contrast, starting from the single-modality baselines and modestly fine tuning higher convolutional layers and the classification head on the target domain can significantly restore-cross device, cross-modal detection performance, with positive class precision and recall approaching or even surpassing the single-domain baselines. Class activation map analysis further indicates that, under fine-tuned and joint training configurations, the models consistently focus on the antenna, the arc, and its surrounding scattering region, supporting physically meaningful explanations and suggesting that cross-device, cross-modal training can capture arc related structures that generalize across devices while providing a lightweight, scalable route to reusing imaging-based diagnostic
Related
- Heterogeneous integrated Pb(Zr, Ti)O₃ thin-film electro-optic modulators via micro-transfer printing at visible and near-infrared wavelengths
- Time-resolved infrared spectroscopy with entangled photons and visible-light detection
- IV-PAFNet: Infrared-visible cross-modal alignment and fusion network for nighttime UAV object detection
- Multi-level registration method for visible and infrared images guided by common optical axis priors
- Object Detection Using Quantum Transfer Learning
BibTeX
@article{guan2026crossdevice,
title = {{Cross‑device and cross‑modal arc detection: Transfer learning between EAST visible and WEST infrared diagnostics}},
author = {Zhongfang Guan and Erwan Grelier and Bin Zhang and Sébastien Vives and Baoguo Wang and Victor Moncada and Leo Dubus and Jinping Qian and Kedong Li and Jian Liu and Cong Cao and Zuhao Wang and Binfu Gao and Chuannan Xuan and Yang Wang and Xiaoming Zhong and Runze Chen and Yutong Guo and Qian Wang and Tianqi Jia and Pan Li and Wenbin Liu and Xiushuo Zhang and Yunchan Hu and Yifan He and Wenyi Lu and Chun Yu He and Xianzu Gong},
journal = {Nuclear Fusion},
year = {2026},
month = oct,
doi = {10.1088/1741-4326/aeb036},
url = {https://cea.hal.science/cea-05787077v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Cross‑device and cross‑modal arc detection: Transfer learning between EAST visible and WEST infrared diagnostics
AU - Zhongfang Guan
AU - Erwan Grelier
AU - Bin Zhang
AU - Sébastien Vives
AU - Baoguo Wang
AU - Victor Moncada
AU - Leo Dubus
AU - Jinping Qian
AU - Kedong Li
AU - Jian Liu
AU - Cong Cao
AU - Zuhao Wang
AU - Binfu Gao
AU - Chuannan Xuan
AU - Yang Wang
AU - Xiaoming Zhong
AU - Runze Chen
AU - Yutong Guo
AU - Qian Wang
AU - Tianqi Jia
AU - Pan Li
AU - Wenbin Liu
AU - Xiushuo Zhang
AU - Yunchan Hu
AU - Yifan He
AU - Wenyi Lu
AU - Chun Yu He
AU - Xianzu Gong
PY - 2026
DA - 2026/10/05
JO - Nuclear Fusion
DO - 10.1088/1741-4326/aeb036
UR - https://cea.hal.science/cea-05787077v1
AB - This work presents an engineering-oriented data processing and model training pipeline for EAST visible and WEST infrared data, enabling a systematic study of cross-device and cross-modal electric arc detection. A lightweight convolutional neural network is adopted as a cross-device arc detection model, and joint training and transfer learning experiments are conducted on EAST visible (VIS) and WEST infrared (IR) data. Single-modality baselines are first trained on EAST and WEST, then evaluated zero-shot on the opposite device, followed by a comparison of frozen-backbone and fine-tuned transfer for VIS→IR and IR→VIS.The results show that models trained from scratch on each diagnostic channel already achieve high precision and recall for arc detection, whereas direct zero-shot cross-modal transfer drives the arc recall on the opposite domain almost to zero. In contrast, starting from the single-modality baselines and modestly fine tuning higher convolutional layers and the classification head on the target domain can significantly restore-cross device, cross-modal detection performance, with positive class precision and recall approaching or even surpassing the single-domain baselines. Class activation map analysis further indicates that, under fine-tuned and joint training configurations, the models consistently focus on the antenna, the arc, and its surrounding scattering region, supporting physically meaningful explanations and suggesting that cross-device, cross-modal training can capture arc related structures that generalize across devices while providing a lightweight, scalable route to reusing imaging-based diagnostic
SN - 0029-5515
SN - 1741-4326
ER -