三维场景感知

三维场景感知是计算机科学和计算机视觉领域中的一个关键方向，旨在使计算机系统能够更全面、准确地理解和解释三维环境。这一领域的目标是模拟人类对于周围世界的感知能力，使计算机能够感知和理解现实世界中的三维场景。

在该领域，我们的研究聚焦于场景生成和补全方向。

（1）三维补全。三维补全从场景中物体的语义和几何出发，学习场景丰富的上下文信息，通过对场景中缺失部分的推断和填充，获得高质量三维场景。

① Qing Guo, Zhijie Wang, Lubo Wang, Haotian Dong, Felix Juefei-Xu, Di Li, Lei Ma, Wei Feng, Yang Liu. CarveNet: Carving Point-Block for Complex 3D Shape Completion. IEEE Transactions on Multimedia (TMM), 2024.

② Haotian Dong, Enhui Ma, Lubo Wang, Miaohui Wang, Wuyuan Xie, Qing Guo, Ping Li, Lingyu Liang, Kairui Yang, Di Lin. CVSformer: Cross-View Synthesis Transformer for Semantic Scene Completion. IEEE International Conference on Computer Vision (ICCV), 2023.

③ Xuzhi Wang, Di Lin (Corresponding Author), Liang Wan. FFNet: Frequency Fusion Network for Semantic Scene Completion. AAAI Conference on Artificial Intelligence (AAAI), 2022.

图1：三维补全介绍

（2）场景生成。场景生成通过自然语言驱动大模型生成自动驾驶仿真场景，以及借助扩散模型等生成模型，生成多视角场景，为下游任务提供更多数据。

④ Kairui Yang, Zihao Guo, Gengjie Lin, Haotian Dong, Die Zuo, Jibin Peng, Zhao Huang, Zhecheng Xu, Fupeng Li, Ziyun Bai, Di Lin. Natural-language-driven Simulation Benchmark and Copilot for Efficient Production of Object Interactions in Virtual Road Scenes. arXiv preprint, 2023.

⑤ Kairui Yang, Enhui Ma, Jibin Peng, Qing Guo, Di Lin, Kaicheng Yu. BEVControl: Accurately Controlling Street-view Elements with Multi-perspective Consistency via BEV Sketch Layout. arXiv preprint, 2023.

⑥ Tingliang Feng, Hao Shi, Xueyang Liu, Wei Feng, Liang Wan, Yanlin Zhou, Di Lin. Open Compound Domain Adaptation with Object Style Compensation for Semantic Segmentation. Neural Information Processing Systems (NeurIPS), 2023.

⑦ Xuzhi Wang, Liang Wan, Di Lin, Wei Feng. Phase-based Fine-grained Change Detection. Expert Systems with Applications, 2023.

⑧ Di Lin, Xin Wang, Jia Shen, Renjie Zhang, Ruonan Liu, Miaohui Wang, Wuyuan Xie, Qing Guo, Ping Li. Generative Status Estimation and Information Decoupling for Image Rain Removal. Neural Information Processing Systems (NeurIPS), 2022.

图2：场景生成介绍