2 Hits in 1.5 sec

ScanQA: 3D Question Answering for Spatial Scene Understanding [article]

Daichi Azuma, Taiki Miyanishi, Shuhei Kurita, Motoaki Kawanabe
2022 arXiv   pre-print
We propose a new 3D spatial understanding task of 3D Question Answering (3D-QA).  ...  Our new ScanQA dataset contains over 40K question-answer pairs from the 800 indoor scenes drawn from the ScanNet dataset.  ...  We consider that 3D spatial understanding datasets contribute to developing models that comprehend the embodied 3D scene and ask and answer questions about the 3D environment as humans do.  ... 
arXiv:2112.10482v3 fatcat:vszty47ybfhabhgyc5bhn5rnhe

3D Question Answering [article]

Shuquan Ye and Dongdong Chen and Songfang Han and Jing Liao
2021 arXiv   pre-print
To verify the effectiveness of our proposed 3DQA framework, we further develop the first 3DQA dataset "ScanQA", which builds on the ScanNet dataset and contains ∼6K questions, ∼30K answers for 806 scenes  ...  Different from image based VQA, 3D Question Answering (3DQA) takes the color point cloud as input and requires both appearance and 3D geometry comprehension ability to answer the 3D-related questions.  ...  In order to answer questions about a real-world 3D scene, 3DQA needs to understand both appearance and 3D geometry.  ... 
arXiv:2112.08359v1 fatcat:gupa6bvijngyvlud4keye3w7f4