Uniworld Semantic Vision

Combine semantic encoders from multimodal LLMs with contrastive learning to create unified high-resolution encoders for both visual understanding and generation tasks without relying on VAE compression.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit aa57069424

Frequently asked questions

npx get-skill add adu2021/uniworld-semantic-vision