Author name cluster

Xiaorui Peng

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

5 papers

2 author rows

EAAI Journal 2026 Journal Article

Text-guided class-incremental point cloud semantic segmentation with category distribution constraint

Chao Zheng
Yan Xu
Xiaorui Peng
Meijun Wang
Yu Meng

Class-incremental light detection and ranging (LiDAR) semantic segmentation refers to maintaining recognition performance on previously learned categories while gradually introducing new categories during learning. This task is particularly challenging due to the sparse, irregular nature of point clouds and the severe class imbalance in large-scale scenes. Existing methods mainly rely on single-modality point-cloud features, which provide limited semantic priors and are vulnerable to category distribution drift across incremental steps. Moreover, its global knowledge distillation tends to make the features of old and novel classes overlap, resulting in confusion between similar categories. In this paper, we propose a novel incremental learning method that effectively integrates multimodal information to enhance the distinction between categories. We introduce textual information and deeply couple the incremental point cloud with text features through a cross-modal attention mechanism, effectively achieving the semantic complementarity across modalities. To further enhance the discriminability of previously learned category boundaries, we propose a historical multi-prototype distribution consistency constraint. In addition, we propose an orthogonality constraint between novel and previous knowledge to mitigate the performance degradation caused by conflicts introduced when learning new categories incrementally. Extensive experiments on the autonomous driving dataset show that our proposed method achieves the state-of-the-art results.

Details DOI

EAAI Journal 2025 Journal Article

Integrating convolutional guidance and Transformer fusion with Markov Random Fields smoothing for monocular depth estimation

Xiaorui Peng
Yu Meng
Boqiang Shi
Chao Zheng
Meijun Wang

Monocular depth estimation is a challenging and prominent problem in current computer vision research and is widely used in intelligent transportation like environment perception, navigation and localization. Accurately delineating object boundaries and ensuring smooth transitions in estimated depth images from a single image remain significant challenges. These issues place higher demands on the network’s global and local feature extraction capabilities. In response, we proposed a depth estimation framework, designed to address detection accuracy and the global smooth transition of predicted depth maps. Our method introduces a novel feature decoding structure named Convolutional Guided Fusion (CoGF), which utilizes local features extracted by a convolutional neural network as a guide and fuses them with long-range dependent features extracted by a Transformer. This approach enables the model to retain both local details and global contextual information during the decoding process. To ensure global smoothness in the depth estimation results, we incorporate a smoothing strategy based on Markov Random Fields (MRF), enhancing pixel-to-pixel continuity and ensuring robust spatial consistency in the generated depth maps. Our proposed method is evaluated on current mainstream benchmarks. Experimental results demonstrate that our depth estimation method outperforms previous approaches. The code is available at https: //github. com/pxrw/CGTF-Depth. git.

Details DOI

ICLR Conference 2025 Conference Paper

Noise Separation guided Candidate Label Reconstruction for Noisy Partial Label Learning

Xiaorui Peng
Yuheng Jia
Fuchao Yang
Ran Wang 0001
Min-Ling Zhang

Partial label learning is a weakly supervised learning problem in which an instance is annotated with a set of candidate labels, among which only one is the correct label. However, in practice the correct label is not always in the candidate label set, leading to the noisy partial label learning (NPLL) problem. In this paper, we theoretically prove that the generalization error of the classifier constructed under NPLL paradigm is bounded by the noise rate and the average length of the candidate label set. Motivated by the theoretical guide, we propose a novel NPLL framework that can separate the noisy samples from the normal samples to reduce the noise rate and reconstruct the shorter candidate label sets for both of them. Extensive experiments on multiple benchmark datasets confirm the efficacy of the proposed method in addressing NPLL. For example, on CIFAR100 dataset with severe noise, our method improves the classification accuracy of the state-of-the-art one by 11.57%. The code is available at: https://github.com/pruirui/PLRC.

Details

AAAI Conference 2024 Conference Paper

Long-Tailed Partial Label Learning by Head Classifier and Tail Classifier Cooperation

Yuheng Jia
Xiaorui Peng
Ran Wang
Min-Ling Zhang

In partial label learning (PLL), each instance is associated with a set of candidate labels, among which only one is correct. The traditional PLL almost all implicitly assume that the distribution of the classes is balanced. However, in real-world applications, the distribution of the classes is imbalanced or long-tailed, leading to the long-tailed partial label learning problem. The previous methods solve this problem mainly by ameliorating the ability to learn in the tail classes, which will sacrifice the performance of the head classes. While keeping the performance of the head classes may degrade the performance of the tail classes. Therefore, in this paper, we construct two classifiers, i.e., a head classifier for keeping the performance of dominant classes and a tail classifier for improving the performance of the tail classes. Then, we propose a classifier weight estimation module to automatically estimate the shot belongingness (head class or tail class) of the samples and allocate the weights for the head classifier and tail classifier when making prediction. This cooperation improves the prediction ability for both the head classes and the tail classes. The experiments on the benchmarks demonstrate the proposed approach improves the accuracy of the SOTA methods by a substantial margin. Code and data are available at: https://github.com/pruirui/HTC-LTPLL.

PDF Details DOI

NeurIPS Conference 2024 Conference Paper

Pedestrian-Centric 3D Pre-collision Pose and Shape Estimation from Dashcam Perspective

Meijun Wang
Yu Meng
Zhongwei Qiu
Chao Zheng
Yan Xu
Xiaorui Peng
Jian Gao

Pedestrian pre-collision pose is one of the key factors to determine the degree of pedestrian-vehicle injury in collision. Human pose estimation algorithm is an effective method to estimate pedestrian emergency pose from accident video. However, the pose estimation model trained by the existing daily human pose datasets has poor robustness under specific poses such as pedestrian pre-collision pose, and it is difficult to obtain human pose datasets in the wild scenes, especially lacking scarce data such as pedestrian pre-collision pose in traffic scenes. In this paper, we collect pedestrian-vehicle collision pose from the dashcam perspective of dashcam and construct the first Pedestrian-Vehicle Collision Pose dataset (PVCP) in a semi-automatic way, including 40k+ accident frames and 20K+ pedestrian pre-collision pose annotation (2D, 3D, Mesh). Further, we construct a Pedestrian Pre-collision Pose Estimation Network (PPSENet) to estimate the collision pose and shape sequence of pedestrians from pedestrian-vehicle accident videos. The PPSENet first estimates the 2D pose from the image (Image to Pose, ITP) and then lifts the 2D pose to 3D mesh (Pose to Mesh, PTM). Due to the small size of the dataset, we introduce a pre-training model that learns the human pose prior on a large number of pose datasets, and use iterative regression to estimate the pre-collision pose and shape of pedestrians. Further, we classify the pre-collision pose sequence and introduce pose class loss, which achieves the best accuracy compared with the existing relevant \textit{state-of-the-art} methods. Code and data are available for research at https: //github. com/wmj142326/PVCP.

PDF Details DOI