search

Mohamedelrefaie / DrivAerNet

A Large-Scale Multimodal Car Dataset with Computational Fluid Dynamics Simulations and Deep Learning Benchmarks
MIT License
133 stars 19 forks source link

The output #5

Closed LearnedVector closed 5 months ago

LearnedVector commented 5 months ago

Hello, I just want to better understand this code base.

The model only predicts a single scale value for drag coefficient? Do you have a version that also predicts the velocity and pressure field?

cseander commented 5 months ago

I'm trying to train a GCN model to predict the pressure in each point, so far the results are not very satisfactory but I'm doing some progress and I will try more sophisticated methods for creating the graph eventually.

Mohamedelrefaie commented 5 months ago

Hey,

Yeah, you're right. The current RegDGCNN model is focused on predicting a single scalar value for the drag coefficient from a 3D STL mesh in an end-to-end manner. We're working on a modified version that can handle surface fields predictions like wall shear stresses and pressure, and we'll be publishing it soon. Keep an eye out for that!

Cheers!