Tran Duc Dang Trung ; Kang Byeongkeun ; Lee Yeejin

Abstract

Recently, transformer-based techniques incorporating superpoints have becomeprevalent in 3D instance segmentation. However, they often encounter anover-segmentation problem, especially noticeable with large objects.Additionally, unreliable mask predictions stemming from superpoint maskprediction further compound this issue. To address these challenges, we proposea novel framework called MSTA3D. It leverages multi-scale featurerepresentation and introduces a twin-attention mechanism to effectively capturethem. Furthermore, MSTA3D integrates a box query with a box regularizer,offering a complementary spatial constraint alongside semantic queries.Experimental evaluations on ScanNetV2, ScanNet200 and S3DIS datasetsdemonstrate that our approach surpasses state-of-the-art 3D instancesegmentation methods.