Exploring 3D Art and AI: RealFusion, NeRF, and Embedding Techniques
Research and Breakthroughs
- Oxford University researchers introduce a diffusion model called RealFusion that can generate 360-degree reconstructions of objects from an image. [https://www.marktechpost.com/2023/02/28/oxford-university-researchers-introduce-a-diffusion-model-called-realfusion-that-can-generate-360-degree-reconstructions-of-objects-from-an-image/?amp]
- Breakthrough for nerf
NeRF and 3D Art
- Devanshu Tak is a 3D artist who has used NeRF earlier in his work
- Interesting read: https://unsupervisedlearning.substack.com/p/using-large-language-models-effectively. Esp the #4 point around embeddings
Embedding and Prompt Templating
- Embedding was the biggest reason for using Pinecone and Langchain in the Hasura bot
- Pinecone is used as a vector store
- Langchain gives a good abstraction
- Hasura bot uses openAI embedding
- Chroma bundles store and the embedding call into one from what can be told.
Links
The description and link can be mismatched because of extraction errors.
- https://www.marktechpost.com/2023/02/28/oxford-university-researchers-introduce-a-diffusion-model-called-realfusion-that-can-generate-360-degree-reconstructions-of-objects-from-an-image/?amp: Oxford University researchers introduce a diffusion model called RealFusion that can generate 360-degree reconstructions of objects from an image. The message in the same link as the URL is not related to the breakthrough.
- https://unsupervisedlearning.substack.com/p/using-large-language-models-effectively: An interesting read about using large language models effectively, with a specific focus on the #4 point around embeddings. The message also mentions the use of Pinecone and Langchain in the Hasura bot for embedding purposes.