search

cvlab-epfl / segmentation-driven-pose

Segmentation-driven 6D Object Pose Estimation. CVPR 2019.
GNU General Public License v3.0
188 stars 37 forks source link

About Occluded-LINEMOD training data #9

Open alen0216056 opened 5 years ago

alen0216056 commented 5 years ago

Hi, Do you generate synthetic data by rendering the provided models? or only using the central objects in LINEMOD scenes?

For each epoch, how many training images do you use? 20k samples mentioned in the paper?

Thanks.

YinlinHu commented 5 years ago

We only use the central objects in LINEMOD, and generate 20K sample images from the central objects by random combinations.