search

DLR-RM / AugmentedAutoencoder

Official Code: Implicit 3D Orientation Learning for 6D Object Detection from RGB Images
MIT License
340 stars 97 forks source link

Object in Rendering not visible #81

Closed ghost closed 4 years ago

ghost commented 4 years ago

Hello, sorry for bothering but I really appreciate your work and I want to make it work for my use-case. I created a custom model.ply but when i try to train the autoencoder I get the warning

Object in Rendering not visible. Have you scaled the vertices to mm?

I tryed to change the vertex_scale parameter in cfg file but nothing.

Here is the 2 model I wanted to work on

test_1.zip test_2.zip

Let me know what it is wrong. Thanks in advance

ghost commented 4 years ago

Ok, I needed to significantly augment the object dimension. It's just strange that changing the vertex_scale parameter didn't work.

ajal-fh commented 4 years ago

Hi,

The same warning is appearing while training on a custom model. The model dimensions are in m (the highest dimension is around 1m) and as instructed 'VERTEX_SCALE' parameter was set to 1000. Do we have to scale down the model? Thanks in advance

ghost commented 4 years ago

Hello,

I'm using Blender to export the .ply models and dimensions of my objects are around 10-20 meters. Using VERTEX_SCALE = 1 works fine for me.