The core technology
The core technology of Neural Concept Shape are 3D convolutional networks that learn to predict the output of physical simulations or experiments based on the input shape’s geometrical properties.
When relevant, the conception process can even be completely delegated to the machine so as to find the optimal design(s) given a number of predefined objectives.
About Shape Expert
Although deep learning has taken the entire field of computer science by storm, Computer Assisted Design (CAD) and geometry processing still mostly rely on traditional techniques.
To change this, the team of researchers behind Neural Concept developed a 3D deep learning system that is specifically dedicated to the processing of CAD and simulation data.
Train your AI once and use it forever
- Transfer learning between use-cases
- Mix experimental and simulation data
Accuracy is key
- Our geometric Neural Networks operate in the euclidean and naturally learn the patterns of physics.
- Analyse performance to improve accuracy and validate models before you use them.
Works with your everyday files
- Geometries: case, vtk, cgns, cas, op2, csv etc...
- Simulation results in all common formats
For advanced experts
The power of deep learning rethought for advanced CAE or optimization experts.
Unique algorithms, taken out of the lab, optimized and made ready to work
- Geometric and 3D based Deep-learning
- Deep-learning from CAD
- Deep Bayesian uncertainty estimation
- Transfer / few-shot learning
- 3D Generative Neural Networks
- Differentiable morphing
- Differentiable CAD
- Online optimization - active learning
- Genetic and gradient - based shape optimization
How it works
Low-level Python-based interface that lets you interact with the core of the technology and removes any limitations
A fully guided workflow that helps you start with the best practices right away.
CAD tool integrations
Miniswys uses Neural Concept Shape for the design optimization of customized ultrasonic actuators