VLM2Vec-V2: Unified Multimodal Embedding for Images, Videos & Documents
VLM2Vec-V2: A unified framework for multimodal embedding learning across images, videos, and documents. Outperforms existing models on various benchmarks.
-
-
Semantic Video Search with Open-Source LLMs on AWS
Revolutionize video search! Use open-source LLMs on AWS for semantic video search via text or image queries. Learn about the architecture, code, and limitations.
-
AI Image Repair: CompleteMe Restores and Edits Human Images
CompleteMe AI excels at restoring and editing human images using reference images, outperforming existing methods in benchmarks. Learn about its features and limitations.
-
MirrorVerse: Revolutionizing AI Reflection Rendering
MirrorVerse improves AI reflection rendering in diffusion models. A new dataset & model, MirrorFusion 2.0, achieves superior results in tests, addressing limitations of previous approaches.
