Congress Final Program

Session

ThS5: Large Language Models for Intelligent LiDAR Point Cloud Processing

Time:

Wednesday, 08-July-2026:

8:30am - 10:00am

Session Chair: Yiping Chen
Session Chair: Jonathan Li

Location: 716B

175 theatre

Session Topics:

Large Language Models for Intelligent LiDAR Point Cloud Processing (ThS5)

Presentations

8:30am - 8:45am

GeoOpen3D: Geometry-guided training-free open-vocabulary 3D segmentation via visual foundation models

Shuai Zhang, Zhuoxiao Li, Jing Ou, Tengxi Wang, Zhecheng Shi, Wufan Zhao

8:45am - 9:00am

SPARC: Scalable 3D Panoptic Segmentation with Reinforcement-driven Clustering

Haoling Huang, Ting Han, Jie Ouyang, Yiping Chen

9:00am - 9:15am

LaSA-Net: A Language-Guided Network for Large-Scale 3D Referring Expression Segmentation on the UrbanRefer Benchmark

Bin Liu, Lingfei Ma, Haiyan Guan, Jonathan Li

9:15am - 9:30am

Scenereasoner: Decoupled Spatial Tokenization for large-scene understanding with llms

Bohang Shi, Shengjun Tang, Xiaoming Li, Weixi Wang, Linfu Xie, Baoding Zhou

9:30am - 9:45am

Llm-Supervised Point Cloud Processing: from Unsupervised 3D Scene-Graph Generation to Interactive Scene Manipulation

Florent Poux, Alex Key

9:45am - 10:00am

Multimodal Large Language Models to road inventory with non-photorealistic Point Cloud visualization

Horia Ameen, Mario Soilán, Henrique Lorenzo Cimadevila, Jesús Balado-Frías