Reconstructed Slides: From Text to Vision to Voice Exploring Multimodality with Open AI: Romain Huet

Source Video

From Text to Vision to Voice Exploring Multimodality with Open AI: Romain Huet

Method

This deck is reconstructed from the existing video frame captures by detecting likely slide regions with OpenCV, cropping/upscaling those regions, deduplicating similar crops, and OCRing the cropped slide images locally. It is a cleaner companion to the full-stage frame deck.

Reconstructed Slides

slide-001.jpg

Dense Scene-Detected Slide Candidates