CalHacks_Pattern_Discovery.ipynb - Colab

This notebook shows how to discover patterns across vision and language using CLIP embeddings and optimal transport. Everything auto-downloads - just run the cells! We'll download diverse images from Unsplash to demonstrate pattern discovery across different visual concepts. Loading... CLIP's power comes from compositional understanding - it can match complex concepts without explicit training. Build a mini image search engine that handles natural language queries. Now let's use optimal transport to understand the "semantic gap" between vision and language. Let's build practical systems that could be used in Conway's pattern discovery pipeline. This notebook demonstrated: These techniques directly support automatic pattern discovery: The embedding space geometry provides a foundation for variety management and dimension reduction in multi-modal contexts.

saved by

• 5914