We’re excited to be heading to this year’s International Conference on Computer Vision (ICCV), taking place October 19–23 in Honolulu, Hawaii. Adobe Research will be showcasing a diverse lineup of research publications that push the boundaries of computer vision and creative technologies.
Our research this year spans a wide spectrum of breakthroughs, including image generation and understanding, 3D reconstruction, video generation and editing, content authenticity, and more. This diverse body of work reflects our mission to empower creators, build trust, and advance the science behind tomorrow’s intelligent systems.
This highlight video offers a closer look at three of our featured innovations:
- Refer to Anything with Vision-Language Prompts — enabling advanced editing by combining text and visual cues to select objects with complex instructions.
- Decomposing the Visual Capabilities of Diffusion Models — uncovering hidden “creative sliders” inside AI image generators, giving users intuitive controls to fine-tune images in novel ways.
- Learning Generalizable Single Image 3D Face Reconstruction from Synthetic Heads — achieving high-quality, identity-preserving 3D head reconstruction from a single photo or video, trained entirely on synthetic data.
These are just a glimpse of the innovations we’re bringing to ICCV 2025. Explore highlights from our publications on our website and see how Adobe Research is advancing both the science and real-world applications of computer vision.
