[CMI2IT] Workshop on Cross-Modal Image-to-Image Translation: Techniques, Applications and Challenges

About the Workshop

This workshop focuses on cross-modal image-to-image translation, where one sensing modality is synthesized from another (e.g., visible–infrared, visible–SAR, RGB–depth, CT–MRI). It welcomes contributions across the full research pipeline, including model architectures, learning under limited or unpaired data, image quality assessment and enhancement, physics- and sensor-aware constraints, and studies of downstream tasks and applications enabled by cross-modal translation. Representative applications include remote sensing and Earth observation, autonomous driving and intelligent transportation, surveillance and security, robotics and unmanned systems, medical imaging, agriculture and environmental monitoring, industrial inspection, and night-vision enhancement.

Call for Papers

Cross-modal image-to-image translation aims to synthesize one sensing modality from another (e.g., visible–infrared, visible–SAR, RGB–depth, CT–MRI), enabling robust perception under adverse conditions and across heterogeneous sensors. Recent advances in diffusion models, GANs, and other generative frameworks have significantly improved the fidelity and diversity of generated images, yet important challenges remain: enforcing physical and sensor consistency, coping with limited or unpaired data, and meeting real-time and reliability requirements in safety-critical scenarios.

This workshop seeks original, unpublished contributions on all aspects of cross-modal image-to-image translation, spanning model design, learning paradigms, evaluation, and applications across different modalities and domains. We particularly welcome work that leverages modern generative models (e.g., diffusion models, GANs, autoregressive or flow-based models) for cross-modal translation in real-world settings. Topics of interest include, but are not limited to:

Architectures for cross-modal image-to-image translation
Low-data / few-shot / unpaired cross-modal learning
Physics-, sensor-, and geometry-aware constraints for physically consistent cross-modal generation
Learning from synthetic data, simulators, or hybrid real–synthetic pipelines
Quality assessment and perceptual / physical metrics for cross-modal generated images
Uncertainty estimation, robustness, and reliability analysis of cross-modal generative models
Real-time and resource-efficient training and inference for cross-modal generators
Downstream tasks enabled by cross-modal translation (e.g., detection, tracking, recognition, segmentation)
Benchmarks, datasets, and evaluation protocols for cross-modal image-to-image translation
Real-world applications and deployment case studies in domains such as remote sensing, autonomous driving, medical imaging, and industrial inspection

Paper Submission

We invite submissions of original research papers addressing but not limited to the topics as listed above. Submissions should adhere to the IEEE ICME 2026 formatting guidelines and will undergo a rigorous peer-review process. Accepted papers will be presented at the workshop and included in the workshop proceeding of the conference*. We also welcome submissions of demos, datasets, and position papers that contribute to the workshop's themes.

Papers have to be submitted via: https://cmt3.research.microsoft.com/IEEEICMEW2026.

Author information and submission instructions are available here: https://2026.ieeeicme.org/author-information-and-submission-instructions/

*All accepted papers need to pay the register fees.

Program

Date and time: July 5, 2026, 2:00 PM – 5:00 PM

Location: Thai Chakkraphat 1

Session	Title	Speaker(s) / Author(s)	Time	Duration
Opening	Opening Remarks	Workshop Chair(s)	2:00–2:05 PM	5 min
Keynote	Self-Perception Network Structure Optimization	Prof. Jiao Shi (Northwestern Polytechnical University)	2:05–2:45 PM	40 min
Keynote	From Metameric Dilemma to Trustworthy RGB-to-HSI	Dr. Xingxing Yang (Hong Kong Baptist University)	2:45–3:25 PM	40 min
Break (3:25–3:30 PM, 5 min)
Oral Presentations
1	Multi-modal Conditioned Bridge Diffusion for Pathology Image Synthesis and Augmentation	Yanan Zhang; Xiangzhi Bai	3:30–3:45 PM	15 min
2	SCISR-TTA: Test Time Adaptation for Screen Content Image Super Resolution	Ajeet Kumar Verma; Vinit Jakhetiya; Badri N Subudhi	3:45–4:00 PM	15 min
3	Diverse Optical-to-SAR Image Translation with Enhanced Feature Disentanglement	Zixiang Ye; Zonghao Han; Kun Hu; Yongping Zhai; Shaohui Mei	4:00–4:15 PM	15 min
4	Visible-to-Infrared Image Translation via Hierarchical Semantic Embeddings	Fangqi Shi; Zonghao Han; Zixiang Ye; Zhengtao Chen; Kun Hu; Shaohui Mei	4:15–4:30 PM	15 min
5	AD-Relight: Training-Free Banner Relighting via Illumination Translation with Diffusion Priors	Rameshwar Mishra; A V Subramanyam	4:30–4:45 PM	15 min
6	CATS-Diff: Channel-Adaptive Timestep Selection for Efficient Diffusion Model Training	Muhammad Ridha Agam; Mas Nurul Achmadiah; Hui-Kai Su; Chi-Chia Sun; Jun-Wei Hsieh	4:45–5:00 PM	15 min