not this specifically

but this

your model seems finished

yep

is still robotic sounds

i think 160 Voice Files still not enough

Hi guys I'm working on a project and was wondering what the best model + setup is for convolutional networks for point clouds?

For example, what would you use for 3D object classificiation?

You can try instant-ngp by NVIDIA

OH that explains a lot How do I know if my model has trained to the right loss then?

I guess only look at the graphs?

This is actually really interesting

Indeed

I say give it a try

I'm retraining my Prime update in Kaggle since Colab doesn't give me enough GPU time.

I hope it's alright to not be a Colab Pro user.

Of course it is

Ok so I spent some time reading the paper and it seems like the first aspect with taking images and creating a 3d flying camera wouldn't be useful... but I'm assuming you were thinking about its ability to take 3d points and create a geometry...? Could you elaborate a bit on what they did there?

watch these

I can share few papers that might be useful for your project

I'll dm them

Thank you so much

in contrary

is having bigger dataset is better?

like having more than 500 voice files?

it definitely won't hurt, but after some point you won't notice the difference

Definitely! 👍