Donghai Zhai

Video class-incremental learning (VCIL) aims to progressively recognize novel action categories while preserving spatial–temporal knowledge of previous tasks. Unlike image class-incremental learning (CIL), VCIL requires simultaneously capturing spatial semantics and temporal dynamics, which makes catastrophic forgetting more significant. Prompt-based learning has recently made remarkable advancements in VCIL. However, they primarily optimize a predefined static prompt for all action sequences in a group-level manner, which overlooks the diverse spatial–temporal characteristics across frames and exhibits limited generalization capability for future action categories. To address the above limitations, we propose a novel exemplar-free VCIL framework termed STCP-low-rank dynamic routing (LRDR) that consists of two crucial innovations, i.e., the spatiotemporal context-aware prompting (SCAP) and LRDR. The former dynamically generates instance-level prompts based on the input video. Specifically, the frame-level prompt is developed to adaptively emphasize fine-grained details in each frame by leveraging the attention-guided spatial activation module. Meanwhile, we also design the cross-frame prompt to capture the differential importance of sequences, allowing the model to focus on key frames and facilitate its ability to explore temporal dependencies. Furthermore, we introduce a parameter-efficient LRDR that achieves dynamic scalability by incorporating mixture-of-experts low-rank spatial–temporal adapters, which can maintain old prompt knowledge and cross-task collaboration as the trainable SCAP also suffers from catastrophic forgetting when learning from incremental tasks. Finally, we propose a novel prompt correction mechanism (PCM) that prevents the proposed SCAP from acquiring ineffective class-wise spatiotemporal representations via the discriminative class-level prompt knowledge. Extensive experiments are conducted on four video benchmarks, and our approach consistently achieved substantial gains over state-of-the-art methods in VCIL.

Monocular vision systems have emerged as a promising solution for 3D object detection due to their cost-effectiveness and deployment simplicity. However, existing methods heavily rely on prior information such as camera parameters and complex 3D dataset annotations. Most current approaches focus on static RGB images, struggling to recover 3D information from 2D inputs. We propose a novel monocular 3D vehicle detection method based on point tracking from a roadside perspective, leveraging dynamic inter-frame associations in video data. Our method operates without prior data like vehicle dimensions or camera parameters, and eliminates the need for complex 3D dataset construction and annotation. It integrates object detection and semantic segmentation for feature extraction, high-precision inter-frame data association, and trajectory-based state analysis. 3D reconstruction is achieved through reference point determination and coordinate calibration under geometric constraints. Experimental results on the DAIR-V2X-I dataset demonstrate superior performance over baseline models, with error reductions ranging from 1.3% to 7.3% across various metrics (A3DS, A2DS, AGS, ALS, APS, AW). This approach presents a practical solution for monocular 3D object detection while eliminating dependency on prior information and complex datasets.

Pavement crack detection is a practical and challenging task that has the ability to significantly reduce the burden of manual building and road maintenance in intelligent transportation systems. Existing methods mainly focus on addressing common crack diseases and are poor in generalizing to other conditions of crack detection due to diverse environmental factors (e.g., illumination), topology complexity, and intensity in-homogeneity. Moreover, the samples suffer from the severe foreground–background imbalance and the model is easily prone to overfitting on trained anomalies, resulting in unsatisfactory performance. To tackle the aforementioned challenges and achieve high-quality pavement crack detection, we propose an innovative approach termed boundary-aware axial attention network (BAAN), which is composed of multiple position-guided axial attention (PAA) modules in a hierarchical encoder–decoder architecture. Specifically, it learns efficient contextual information via decomposed multidimensional position-guided attention to capture more precise spatial structures, and the proposed boundary regularization module (BRM) mines more discriminative foreground–background relationships to regularize the ambiguous details between diverse spatial regions. Moreover, we propose a novel boundary refinement loss (BRL) to alleviate the challenges associated with regional losses (e.g., pixel-wise cross-entropy loss) in the context of heavily imbalanced crack detection problems. The proposed BAAN is evaluated on four crack datasets and experimental results indicate that the BAAN consistently outperforms the state-of-the-art methods with fewer computational requirements.

There are a huge number of models claim to enhance visual surface defect inspection accuracy. However, as these models generally function directly within the pixel space, optimizing advanced segmentation techniques frequently demands substantial computational resources and poses challenges for inference on devices with limited computing power. Additionally, many current methodologies are deeply reliant on extensive surface defect datasets. In response to these challenges, our research presents a novel approach based on an auto-encoder structure that uses “latent space” to refine defect segmentation. Within the encoder segment of the autoencoder, we’ve incorporated contrastive learning, amplifying both feature extraction and segmentation capabilities. This architectural choice not only tailors the strategy for prompt response scenarios and underscores its precision in high-accuracy applications, but also addresses the challenges posed by the scarcity of defect samples. As a means to assess our approach and better cater to industrial applications that prioritize sample-level accuracy, we introduce innovative sample-level metrics, namely Mostly Segmented (MS) and Mostly Lost (ML). Experiments conducted on the RSDD and Neuseg datasets underscore the strategy’s steadfast performance under diverse data circumstances. Synthesizing the benefits of the latent space and the contrastive learning, this paper delineates proficient methodology for surface defect segmentation.

In our daily lives, people frequently consider daily schedule to meet their needs, such as going to a barbershop for a haircut, then eating in a restaurant, and finally shopping in a supermarket. Reasonable activity location or point-of-interest (POI) and activity sequencing will help people save a lot of time and get better services. In this article, we propose a reinforcement learning-based deep activity factor balancing model to recommend a reasonable daily schedule according to user’s current location and needs. The proposed model consists of a deep activity factor balancing network (DAFB) and a reinforcement learning framework. First, the DAFB is proposed to fuse multiple factors that affect daily schedule recommendation (DSR). Then, a reinforcement learning framework based on policy gradient is used to learn the parameters of the DAFB. Further, on the feature storage based on the matrix method, we compress the feature storage space of the candidate POIs. Finally, the proposed method is compared with seven benchmark methods using two real-world datasets. Experimental results show that the proposed method is adaptive and effective.

Parkinson’s disease (PD) is a neurodegenerative disease of the brain associated with motor symptoms. With the maturation of machine learning (ML), especially deep learning, ML has been used to assist in the diagnosis of PD. In this paper, we explore graph neural networks (GNNs) to implement PD prediction using MRI data. However, most existing GNN models suffer from the efficiency of graph construction on MRI data and the problem of overfitting on small data. This paper proposes a novel multi-layer GNN model that incorporates a fast graph construction method and a sparsity-based pooling layer with an attention mechanism. In addition, graph structure sparsity is plugged into the graph pooling layer as prior knowledge to mitigate overfitting in model training. Experimental results on real-world datasets demonstrate the effectiveness of the proposed model and its superiority over baseline methods.