Leveraging Scene Graphs for Generative AI and Innovative Data Products
summary
Claims that converting images and video into explicit graphs yields stronger conditioning for generative models, reducing illogical outputs (e.g., extra limbs) and enabling controllable image and video synthesis for time of day, weather and surgical simulation. Demonstrations link scene graphs to 3D scene understanding, VQA and cross-modal retrieval, and to productized features: scene graphs as structured conditioning, optimized store layouts, customer behavior graphs across cameras and personalized game environments using Nidia freestyle and tools from Nahdet Misr Digital Solutions. Results report clearer constraints and more predictable generations.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:23Scene graphs basics
- 3:34Graphs improve AI understanding
- 6:41Scene graphs in computer vision
- 10:43Scene graphs for generative AI
- 13:38Personalized media with graphs
- 15:30Data products using scene graphs
- 18:03Scene graphs in traffic systems
related talks
45mGenerative AI for Image and Video Production
15mPractical Evolution of Data Science with GenAI
28mApplications of Generative Modeling
17mThe State of Generative Media
19mSketchforme: Composing Sketched Scenes from Text Descriptions for Interactive Applications
06mData Science
01h 26mThinking in Graphs: The Cognitive Foundations of Agentic AI
34mGenerative AI Landscape Building a Modern Data Strategy