Input images

[manycore-research]

Input images

[bertjiazheng]

We convert SVG files to raster images for training using the code available at https://github.com/manycore-research/PlankAssembly/blob/atlas/plankassembly/data.py#L200.

Best,
Jia

[guist]

Sorry, I think my explanations were not clear. I already have raster images, I don't need to generate them from svg files.

My question was: how to use raster images as input to your testing pipeline?

[bertjiazheng]

You should pay attention to the intrinsics and pose parameters.

• For orthographic projection, we set the intrinsic matrix as here.
• We set the poses of the front, top, right view in here.

Best,
Jia

[bertjiazheng]

Feel free to reopen this issue if you still have any questions.

[guist]

Hello,

Sorry for the late answer.

Thanks for your precisions. I tried to modify the code as much as possible but unfortunately I have not managed to make it run with my own data. Therefore I am going to try with a different model.

Thanks again for your help!

Best,
Guillaume