GALA is built on an LLM backbone with a point cloud encoder and a projector that maps geometry features into the language embedding space.
Point cloud aware language model enables controllable fine-grained 3D object placement
GALA predicts 6D poses one object at a time, using explicit geometry to handle support, containment and collision constraints that bounding boxes miss.
Chinese Tech
Chucheng Xiang · Runze Wang · Zhi Deng · Ruchao Bao · Yuanwei Zhang · Yuxuan Xie · +7 more
University of Science and Technology of China · Tencent · Tsinghua University · Hong Kong University of Science and Technology
Research Digest··2 min read
The authors introduce GALA, a multimodal language model that consumes point clouds to predict 6D object poses for 3D scene arrangement.
§