arXiv preprint · cs.CV
Image-Space Rule Discovery
Introduces WISRD, a benchmark for testing whether image-editing models can interpret visual instructions, infer rules, and complete worksheet-style tasks directly in image space.
Research profile
Misora Sugiyama
杉山 未空
Computer Vision · Image & Video Generation
02 — Publications
Selected publications
Two public arXiv preprints spanning computer vision, generative modeling, image editing, and visual reasoning.
View on arXiv
arXiv · ICCV 2025 Workshop
Simple Visual Artifact Detection in Sora-Generated Videos
A multi-label framework for detecting four common types of visual artifacts in Sora-generated videos, achieving 94.14% average classification accuracy with ResNet-50.
03 — What's new?
What's new?
04 — Projects
Research projects
Computer Vision
Learning representations that capture objects, structure, context, and change in visual data.
Image & Video Generation
Studying generative models for coherent, controllable, and expressive visual synthesis.
Image-to-Image Models
Translating images across domains while preserving the content and intent that matter.
05 — Bio
Bio.
My work sits at the intersection of computer vision and generative AI. I am interested in models that do more than create convincing pixels—models that understand structure, preserve intent, and translate visual information across domains.
Current interests include image and video generation, image-to-image models, controllable synthesis, and the representations that connect perception with generation.
06 — Affiliations
Affiliations
NYU Tandon School of Engineering
Academic affiliation
LIMIT Lab
Research community
cvpaper.challenge
Computer vision community
07 — Contact
Contact
For research inquiries, talks, or potential collaborations, feel free to get in touch.