Professor: ‘if it doesn’t look correct because of our models being too off, we will know by seeing it’
How does that work when dynamically generating motion simulations that have never been run before? Is there a more systematic way to know if something is off without rendering it?
ncastaneda02
Generally, you likely wouldn't be able to check if your motion model is absolutely correct without just rendering it, as there really isn't any easy way to tell what your model will do without just simulating it. However, this is why you oftentimes see really coarse subdivision surfaces with no shading during the motion stage of animation as it will render quickly so you can afford to just render the scene to see if you are correct. I'm sure there is also a fancy formal verification/ML model that can prove the correctness of your numerical animation model, but that might be a bit overengineered.
Professor: ‘if it doesn’t look correct because of our models being too off, we will know by seeing it’
How does that work when dynamically generating motion simulations that have never been run before? Is there a more systematic way to know if something is off without rendering it?
Generally, you likely wouldn't be able to check if your motion model is absolutely correct without just rendering it, as there really isn't any easy way to tell what your model will do without just simulating it. However, this is why you oftentimes see really coarse subdivision surfaces with no shading during the motion stage of animation as it will render quickly so you can afford to just render the scene to see if you are correct. I'm sure there is also a fancy formal verification/ML model that can prove the correctness of your numerical animation model, but that might be a bit overengineered.