Use my own 3d model, but it seems not right

DLR-RM / AugmentedAutoencoder

Official Code: Implicit 3D Orientation Learning for 6D Object Detection from RGB Images

MIT License

339 stars 97 forks source link

Use my own 3d model, but it seems not right #17

Closed xiaomaoo closed 5 years ago

xiaomaoo commented 5 years ago

I replaced the "obj_5.ply" model to my own ply model, which has the same format with the model in the models_reconst.However, it looks wrong when I run "ae_train exp_group/my_autoencoder -d" Why?

screenshot from 2019-03-05 20-30-00

MartinSmeyer commented 5 years ago

Might be the scale of the 3D model. Change the VERTEX_SCALE parameter to 1000 in the train cfg.

xiaomaoo commented 5 years ago

Thanks for reply! I tried, but didn’t work. I attached my model in the attachment: mymodel.zip

MartinSmeyer commented 5 years ago

Hi, additionally to the wrong vertex scale, the model origin is kilometers away from the actual origin.

I centered the model origin for you. You can use meshlab for it :) mymodel_centered.ply.zip

With this model the training preview looks fine.

xiaomaoo commented 5 years ago

Thank you very much!