2026.07
"FreeEyeglass: Training-free and Mask-free Eyeglass Transfer for Facial Videos" has been accepted to TMLR.
Weng Ian (Ianna) Chan
PhD Candidate, The University of Osaka
I am a doctoral student at Osaka University, Japan, advised by Prof. Fumio Okura. My research interests lie in computer vision and generative models, with a particular focus on controllable image and video generation, diffusion models, and developing generative technologies that are more controllable, reliable, and practical in real-world applications.
I received both my M.S. in Information Science and B.Eng. from Osaka University. From July to November 2024, I completed a research internship at CyberAgent, Inc., where I worked on facial video editing.
News
Recent updates
2026.04
"V-PartSwap: Motion-Consistent Facial Part Transfer in Videos via Alignment-Aware Diffusion" has been accepted to CVPR Workshop on Video Generative Models: Benchmarks and Evaluation.
2025.10
"Instance-wise Distribution Control of Text-to-image Diffusion Models" appears in Pattern Recognition.
2025.08
"FreeEyeglass: Training-free and Mask-free Eyeglass Transfer for Facial Videos" received the Audience Award and Student Encouragement Award in MIRU 2025.
2024.11 – 2025.03
Visited University of California, Merced, in Prof. Ming-Hsuan Yang's lab.
Projects
Research projects
TMLR
FreeEyeglass: Training-free and Mask-free Eyeglass Transfer for Facial Videos
A training-free, mask-free approach for transferring eyeglasses in facial videos, designed for practical video editing without additional model training or manually prepared masks.
V-PartSwap: Motion-Consistent Facial Part Transfer in Videos via Alignment-Aware Diffusion
CVPRW 2026
V-PartSwap
A training-free approach to facial part transfer in video using DiffAE.
Instance-wise Distribution Control of Text-to-image Diffusion Models
Pattern Recognition, 2026
Instance-wise Distribution Control of Text-to-image Diffusion Models
A method for controlling multi-instance distributions in text-to-image diffusion models, with the goal of improving controllability and representation of underrepresented groups.
Quantifying the recovery process of skeletal muscle on hematoxylin and eosin stained images via learning from label proportion
Scientific Reports, 2024
Quantifying the recovery process of skeletal muscle on hematoxylin and eosin-stained images via learning from label proportion
A weakly supervised medical-imaging project that quantifies skeletal muscle recovery from hematoxylin and eosin-stained images using learning from label proportion.
Learning from Similarity Proportion Loss for Classifying Skeletal Muscle Recovery Stages
MICCAI ADS-MIA Workshop, 2024
Learning from Similarity Proportion Loss for Classifying Skeletal Muscle Recovery Stages
A label-proportion-based learning approach for classifying skeletal muscle recovery stages, combining weak supervision with medical imaging analysis.
MIRU 2023 · Interactive Presentation Award
Fine-grained Facial Image Manipulation via Latent Space Decomposition
A latent-space decomposition approach for fine-grained facial image manipulation, enabling more detailed control of facial attributes and expressions.
MIRU 2021
Conditional StyleGAN with Multi-resolution Classifiers for Attribute-driven Face Synthesis
A StyleGAN-based face synthesis project using multi-resolution classifiers for attribute-driven generation and control.
For a complete record, see the full publication list.
Contact
Get in touch
chan.wengian{at}ist.osaka-u.ac.jp
(replace {at} with @)
Affiliation
Graduate School of Information Science and Technology
The University of Osaka
1-5 Yamadaoka, Suita, Osaka 565-0871, Japan