#SceneStitch
summary
Image hole-filling system uses Adobe Sensei to query Adobe Stock and pull patches from 100 million images instead of only from the same photo. It returns multiple candidate fills, uses semantic matching not just pixel similarity, and can replace roads, foreground parks or whole neighborhoods with alternatives like rivers, fields or beach scenes. Demos produce many ready-to-use results but some mismatched reflections or context appear, so selection and minor tweaks remain necessary.
AI-written summary · about
related talks
14mLeveraging Artificial Intelligence for Scene Detection in Broadcast Video
01h 15mPre-sets and Post-sets: Poetics and Projections
43mThe Search for a New Visual Search, Beyond Language
26mAI at Stitch Fix
01h 27mImage Generation with AI
36mArt in the Age of Algorithms
22mHarnessing image and text data for product matching
19mAI-Powered Startup Toolkit for a Product Designer