In this paper, we present Change3D, a framework that reconceptualizes the change detection and captioning tasks through video modeling. Recent methods have achieved remarkable success by regarding each pair of bi-temporal images as separate frames. They employ a shared-weight image encoder to extract spatial features and then use a change extractor to capture differences between the two images. However, image feature encoding, being a task-agnostic process, cannot attend to changed regions effectively. Furthermore, different change extractors designed for various change detection and captioning tasks make it difficult to have a unified framework. To tackle these challenges, Change3D regards the bi-temporal images as comprising two frames akin to a tiny video. By integrating learnable perception frames between the bi-temporal images, a video encoder enables the perception frames to interact with the images directly and perceive their differences. Therefore, we can get rid of the intricate change extractors, providing a unified framework for different change detection and captioning tasks. We verify Change3D on multiple tasks, encompassing change detection (including binary change detection, semantic change detection, and building damage assessment) and change captioning, across eight standard benchmarks. Without bells and whistles, this simple yet effective framework can achieve superior performance with an ultra-light video model comprising only ~6%-13% of the parameters and ~8%-34% of the FLOPs compared to state-of-the-art methods. We hope that Change3D could be an alternative to 2D-based models and facilitate future research.

Recent advances in scene understanding have leveraged multimodal large language models (MLLMs) for 3D reasoning by capitalizing on their strong 2D pretraining. However, the lack of explicit 3D data during MLLM pretraining limits 3D representation capability. In this paper, we investigate the 3D-awareness of MLLMs by evaluating multi-view correspondence and reveal a strong positive correlation between the quality of 3D-aware representation and downstream task performance. Motivated by this, we propose 3DRS, a framework that enhances MLLM 3D representation learning by introducing supervision from pretrained 3D foundation models. Our approach aligns MLLM visual features with rich 3D knowledge distilled from 3D models, effectively improving scene understanding. Extensive experiments across multiple benchmarks and MLLMs—including visual grounding, captioning, and question answering—demonstrate consistent performance gains. Project: https://visual-ai.github.io/3drs/

In this paper, we introduce \textbf{FROSTER}, an effective framework for open-vocabulary action recognition. The CLIP model has achieved remarkable success in a range of image-based tasks, benefiting from its strong generalization capability stemming from pretaining on massive image-text pairs. However, applying CLIP directly to the open-vocabulary action recognition task is challenging due to the absence of temporal information in CLIP's pretraining. Further, fine-tuning CLIP on action recognition datasets may lead to overfitting and hinder its generalizability, resulting in unsatisfactory results when dealing with unseen actions. To address these issues, FROSTER employs a residual feature distillation approach to ensure that CLIP retains its generalization capability while effectively adapting to the action recognition task. Specifically, the residual feature distillation treats the frozen CLIP model as a teacher to maintain the generalizability exhibited by the original CLIP and supervises the feature learning for the extraction of video-specific features to bridge the gap between images and videos. Meanwhile, it uses a residual sub-network for feature distillation to reach a balance between the two distinct objectives of learning generalizable and video-specific features. We extensively evaluate FROSTER on open-vocabulary action recognition benchmarks under both base-to-novel and cross-dataset settings. FROSTER consistently achieves state-of-the-art performance on all datasets across the board. Project page: \url{https://visual-ai.github.io/froster}.

Abstract Fault recorded data has been proven to be effective for fault diagnosis of overhead transmission lines. Utilizing deep learning to mine potential fault patterns in fault recording data is an inevitable trend. However, it is usually difficult to obtain massive labeled fault recording data, which results in deep learning‐based fault diagnosis models not being adequately trained. Although data augmentation methods provide ideas for expanding the training data, existing data augmentation algorithms (e.g. random perturbation‐based augmentation) may lead to distortion of multi‐view data, that is, time domain data and frequency domain data of the fault recorded data, which results in the inconsistency of physical properties and statistical distributions of the generated data and the actual recording data, and misguides the training of the models. Hence, this study proposes a transmission line fault classification method via the multi‐view synergistic enhancement of fault recording data. The methodology proposes to start with a synergistic enhancement of multi‐view data such as time and frequency domains of fault recording data, and utilizes contrastive learning to further improve the performance of the fault classification model while ensuring that the generated data is not distorted. Experimental results on three real‐world datasets validate the effectiveness of the proposed method.

Abstract Insulator defect detection is crucial for the stable operation of power systems. It has become a mainstream research direction to realise insulator defect detection based on the combination of line images captured by UAVs and deep learning techniques. However, the existing high‐quality insulator defect detection models still face problems such as relying on massive‐labelled data and huge model parameters. Especially on resource‐constrained devices, it becomes a challenge to strike a balance between model lightweighting and performance. Although the knowledge distillation technique provides a solution for model lightweighting, the loss of information in the distillation process leads to the performance degradation of small models, which in turn creates a paradox between lightweighting and performance. Hence, an insulator defect detection method based on federated knowledge distillation is proposed. The method not only realises the lightweighting of the model, but also effectively improves the model performance by collaboratively training the model through the federated learning approach. Moreover, the asynchronous aggregation approach and model freshness mechanism designed in the method further enhance the training efficiency and collaborative effect. The experimental results show that the detection accuracy and efficiency of this paper's method on public datasets are significantly better than the benchmark algorithm.

Graph convolutional networks have been widely applied in skeleton-based gait recognition. A key challenge in this task is to distinguish the individual walking styles of different subjects across various views. Existing state-of-the-art methods employ uniform convolutions to extract features from diverse sequences and ignore the effects of viewpoint changes. To overcome these limitations, we propose a condition-adaptive graph (CAG) convolution network that can dynamically adapt to the specific attributes of each skeleton sequence and the corresponding view angle. In contrast to using fixed weights for all joints and sequences, we introduce a joint-specific filter learning (JSFL) module in the CAG method, which produces sequence-adaptive filters at the joint level. The adaptive filters capture fine-grained patterns that are unique to each joint, enabling the extraction of diverse spatial-temporal information about body parts. Additionally, we design a view-adaptive topology learning (VATL) module that generates adaptive graph topologies. These graph topologies are used to correlate the joints adaptively according to the specific view conditions. Thus, CAG can simultaneously adjust to various walking styles and viewpoints. Experiments on the two most widely used datasets (i.e., CASIA-B and OU-MVLP) show that CAG surpasses all previous skeleton-based methods. Moreover, the recognition performance can be enhanced by simply combining CAG with appearance-based methods, demonstrating the ability of CAG to provide useful complementary information.

Although gait recognition has drawn increasing research attention recently, it remains challenging to learn discriminative temporal representation since the silhouette differences are quite subtle in spatial domain. Inspired by the observation that humans can distinguish gaits of different subjects by adaptively focusing on temporal sequences with different time scales, we propose a context-sensitive temporal feature learning (CSTL) network in this paper, which aggregates temporal features in three scales to obtain motion representation according to the temporal contextual information. Specifically, CSTL introduces relation modeling among multi-scale features to evaluate feature importances, based on which network adaptively enhances more important scale and suppresses less important scale. Besides that, we propose a salient spatial feature learning (SSFL) module to tackle the misalignment problem caused by temporal operation, e.g., temporal convolution. SSFL recombines a frame of salient spatial features by extracting the most discriminative parts across the whole sequence. In this way, we achieve adaptive temporal learning and salient spatial mining simultaneously. Extensive experiments conducted on two datasets demonstrate the state-of-the-art performance. On CASIA-B dataset, we achieve rank-1 accuracies of 98.0%, 95.4% and 87.0% under normal walking, bag-carrying and coat-wearing conditions. On OU-MVLP dataset, we achieve rank-1 accuracy of 90.2%. The source code will be published at https://github.com/OliverHxh/CSTL.

Data in the form of spatial point patterns are frequently encountered in some manufacturing processes such as the nanoparticle reinforced composite materials and defects on semiconductor wafers. Their spatial characteristics contain rich information about the fabrication processes and are often strongly related to the product quality. The distributional characteristics of a spatial point pattern can be summarized by functional profiles such as the popular Ripley's K function. By analyzing the K function, we can effectively monitor the distributional behaviors of the spatial point data. In this study, statistical properties of the K function are investigated, and a Gaussian process is found to be appropriate in characterizing its behavior under complete spatial randomness. A control chart is proposed based on the results to monitor the uniformity of point patterns. Our proposed approach has been compared with existing methods through numerical simulations and shown superior performances.