Step into an AI 3D world
Lying and intuition are traditionally the domain of human. However, this is changing; AI models hallucinate like chatGPT or intuitively model 3D like in this method described below.
🟠 Start creating 3D with chatGPT
In computer vision and graphics, the goal is still to map high-quality 3D geometry to a representation that can be easily changed. The most challenging part is having to model both continuous and discrete changes in shape. Current methods, like implicit shape representation, don’t have clear, easy-to-understand encoding, while procedural methods produce rough geometry.
The new method for making 3D shapes using a parameter space that is easy to change.
The researchers created a new program that enforces a complex set of rules and lets users make high-level edits that are easy to understand and control and that spread procedurally at a low level to the whole shape. By design, the program makes mesh outputs that are of high quality.
The authors use a neural network to map a point cloud or sketch a space with parameters that can be understood. Once a shape has been made by the procedural program, it is easy to change it. They show that GeoCode can infer and recover 3D shapes more accurately than other methods and that it can be used to change shapes in a controlled way both locally and globally.
Step into a world of limitless possibility with AI in 3D art
AI in 3D art is a rapidly evolving field that opens up new possibilities for creative expression and the generation of…
GeoCode Method Overview
GeoCode learns to map a point cloud or sketch to a parameter space that is easy to change. The input goes through the corresponding encoder to get an embedding vector, which is then sent to a set of decoders that predict the interpretable parameters. The program follows a set of rules that builds a good shape when given a representation of parameters.
Input settings that are intuitive
The program works with three human-intuitive input parameters: discrete, binary, and continuous. The program gives you separate control over the shape, so you can change one part while keeping the others the same. But it also models complex structural interactions, where one piece affects another, and the second part adjusts automatically to keep details touching and intact the shape’s structure.
🟠 Start creating 3D with chatGPT
The world of 3D AI has been accelerating at a rapid pace, much like a Formula One race car. In 2023, we should see an…
What seems obvious for a human, e.g. which elements of an object are related to each other, becomes clear for AI models thanks to this method
In the future, the researchers want to add more features to the procedural program, such as UV texture maps and properties of the materials used to make the shapes. They also want to look into ways to improve how AI represents 3D scenes.
New 3D generation method - 2×faster
Gaming, entertainment, architecture, and robotics simulation use 3D digital content. It's spreading to shopping…
I invite you to explore the concept of Creative Machine Learning tools by reading and learning from the many articles found on 🔵 MLearning.ai 🟠
- Check out my instagram with new material every week
- If you enjoyed this, follow me on Medium for more
- Want to collaborate? Let’s connect on LinkedIn
- https://linktr.ee/evartology
Data Scientists must think like an artist when finding a solution when creating a piece of code. Artists enjoy working on interesting problems, even if there is no obvious answer.
All our writers (members) receive the opportunity to be promoted on our social media, which increases the popularity of articles published on MLearning.ai
- Linkedin (44K+ ML-professionals)
- Twitter (5.4K+ followers)
- Instagram (2.4K + followers )
- Sketchfab * — individual vRooML!
- Youtube
- Apple Podcasts
- Substack
Keywords: computer vision, Artificial Intelligence, Machine Learning, AI art, art, wombo dream, digital art, Dalle 2, Imagen, wombo ai, Parti, text-to-image, diffusion models, generative art, wombo art, photographic quality, chatGPT, img by AI system, AI art generator, text to art generator, free ai art generator, 3D ai art, 3D
https://arxiv.org/pdf/2212.11715.pdf
PROJECT PAGE:
https://arxiv.org/pdf/2212.11715.pdf
CODE:
https://github.com/threedle/GeoCode
@article{pearl2022geocode,
title={GeoCode: Interpretable Shape Programs},
author={Pearl, Ofek and Lang, Itai and Hu, Yuhua and Yeh, Raymond A. and Hanocka, Rana},
booktitle={arXiv preprint arxiv:2212.11715},
year={2022}
}