CS 6674

Multimodal Computer Vision

Course Description

Course information provided by the 2026-2027 Catalog.

Multimodal representations are reshaping computer vision, driving advances in both understanding and generation across a wide range of perceptual tasks. This research-oriented course explores computer vision techniques that integrate images with additional modalities such as language and 3D geometry for addressing challenges in both analysis and synthesis tasks. Possible topics include visual grounding, multimodal alignment, and text-guided generation and editing over multiple 2D and 3D representations.

Enrollment Information Enrollment limited to: Cornell Tech Doctor of Philosophy (PhD) students. Recommended prerequisites: CS 3780 or CS 5780; CS 5670; or equivalents.

Last 4 Terms Offered 2026SP

Learning Outcomes

• Analyze state-of-the-art multimodal techniques and understand their architectures and capabilities.
• Identify open challenges in the field of multimodal computer vision.
• Design deep learning pipelines that combine multiple modalities for addressing a research-oriented problem in computer vision.

View Enrollment Information