Point2RBox-v2: Rethinking Point-supervised Oriented Object Detection with Spatial Layout Among Instances

Yi Yu, Botao Ren, Peiyuan Zhang, Mingxin Liu, Junwei Luo, Shaofeng Zhang, Feipeng Da, Junchi Yan, Xue Yang

Project

• This work rethinks point-supervised oriented object detection with the layout among instances. At the core are three principles: 1) Gaussian overlap loss. 2) Voronoi watershed loss. 3) Consistency loss. These principles lead to strong performance.

Pointobb-v3: Expanding Performance Boundaries of Single Point-Supervised Oriented Object Detection

Peiyuan Zhang, Junwei Luo, Xue Yang, Yi Yu, Qingyun Li, Yue Zhou, Xiaosong Jia, Xudong Lu, Jingdong Chen, Xiang Li, Junchi Yan, Yansheng Li

Project

• This work presents an extended conference version of PointOBB, which incorporates a novel Scale-Sensitive Feature Fusion (SSFF) module to improve the model’s capability of perceiving object scales, and further proposes an end-to-end optimized framework.

Envisioning Beyond the Pixels: Benchmarking Reasoning-Informed Visual Editing

Xiangyu Zhao, Peiyuan Zhang, Kexian Tang, Xiaorong Zhu, Hao Li, Wenhao Chai, Zicheng Zhang, Renqiu Xia, Guangtao Zhai, Junchi Yan, Hua Yang, Xue Yang, Haodong Duan

Project

• This paper proposes RISEBench, the first benchmark for reasoning-informed visual editing, covering four core reasoning tasks—Temporal, Causal, Spatial, and Logical—and introducing a comprehensive evaluation framework with three key dimensions: Instruction Reasoning, Appearance Consistency, and Visual Plausibility.

Partial Weakly-Supervised Oriented Object Detection

Mingxin Liu, Peiyuan Zhang, Yuan Liu, Wei Zhang, Yue Zhou, Ning Liao, Ziyang Gong, Junwei Luo, Zhirui Wang, Yi Yu, Xue Yang

Project

• This paper proposes PWOOD, a cost-effective framework for oriented object detection that uses partially weak and unlabeled data through orientation- and scale-aware learning, achieving competitive performance with much lower annotation cost.