search

DLR-RM / AugmentedAutoencoder

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

[question]About the LineMod test data #108

Closed JZhao12 closed 2 years ago

JZhao12 commented 2 years ago

Hi Martin,

Sorry to disturb you, may I ask, did you use all the LineMod data to test AAE? if not, what was the test data when you evaluated AAE based on LineMod?

Many thanks Jianyu

MartinSmeyer commented 2 years ago

Hi Jianyu,

sorry, I totally forgot to answer you. I used the usual 70% split for testing as described in other papers if I remember correctly.

Best,

Martin

JZhao12 commented 2 years ago

No worries Martin, I saw that BOP 2022 just opened and you must be very busy with this.

I already found the answer from Brachmann's paper: " Each object sequence is split in training and test data. Training images are selected such that the associated object poses have a minimum angular distance of 15◦ . Doing so selects ≈ 15% of the images for training. "

Thank you very much for your reply! Jianyu