Closed naaths closed 3 years ago
Hi,
We provide models trained on the linemod synthetic PBR images and evaluated them on the Linemod-Occluded dataset for the BOP challenge. The code for training and evaluation of these models is available in this repository (see section on bop20 challenge). I haven't evaluated these models on Linemod but it is very easy to modify bop20 inference and evaluation scripts to evaluate on this dataset.
Note however that this wouldn't be a fair comparison with most other works that use parts of real linemod images for training whereas our models are only trained on PBR synthetic images. I did not train models on the real linemod images because there is no official training split with real images for Linemod on the BOP website.
I may train and evaluate our approach on the linemod training/test splits commonly used by other works in a future version of the paper.
Hello, may I ask, can this code directly train and evaluate the linemod dataset? Thank you