This submission proposes a method for processing point clouds augmented with incorporating global context features in learning. It initially received four reviews with mixed positive and negative scores (7,5,6,2). The reviewers mention excellent performance, thorough evaluation, and novelty in applying and extending known technics to object detection in 3D. The rebuttal clarified the majority of other concerns, which resulted in an increase of scores to (7,6,6,6). For these reasons, the AC's recommendation is to accept this submission as a poster.