Linlin Zhang

Rui Ying, Mengting Hu, Jianfeng Wu et al.

Temporal knowledge graph completion aims to infer the missing facts in temporal knowledge graphs. Current approaches usually embed factual knowledge into continuous vector space and apply geometric operations to learn potential patterns in temporal knowledge graphs. However, these methods only adopt a single operation, which may have limitations in capturing the complex temporal dynamics present in temporal knowledge graphs. Therefore, we propose a simple but effective method, i.e. TCompoundE, which is specially designed with two geometric operations, including time-specific and relation-specific operations. We provide mathematical proofs to demonstrate the ability of TCompoundE to encode various relation patterns. Experimental results show that our proposed model significantly outperforms existing temporal knowledge graph embedding models. Our code is available at https://github.com/nk-ruiying/TCompoundE.

Yuhang Wang, Zhenxing Niu, Haoxuan Ji et al.

The core challenge of machine unlearning is to strike a balance between target knowledge removal and non-target knowledge retention. In the context of Multimodal Large Language Models (MLLMs), this challenge becomes even more pronounced, as knowledge is further divided into visual and textual modalities that are tightly intertwined. In this paper, we introduce an MLLM unlearning approach that aims to forget target visual knowledge while preserving non-target visual knowledge and all textual knowledge. Specifically, we freeze the LLM backbone and achieve unlearning by fine-tuning the visual module. First, we propose a Contrastive Visual Forgetting (CVF) mechanism to separate target visual knowledge from retained visual knowledge, guiding the representations of target visual concepts toward appropriate regions in the feature space. Second, we identify the null space associated with retained knowledge and constrain the unlearning process within this space, thereby significantly mitigating degradation in knowledge retention. Third, beyond static unlearning scenarios, we extend our approach to continual unlearning, where forgetting requests arrive sequentially. Extensive experiments across diverse benchmarks demonstrate that our approach achieves a strong balance between effective forgetting and robust knowledge retention.

Zhisheng Lu, Juncheng Li, Hong Liu et al.

Single image super-resolution (SISR) has witnessed great strides with the development of deep learning. However, most existing studies focus on building more complex networks with a massive number of layers. Recently, more and more researchers start to explore the application of Transformer in computer vision tasks. However, the heavy computational cost and high GPU memory occupation of the vision Transformer cannot be ignored. In this paper, we propose a novel Efficient Super-Resolution Transformer (ESRT) for SISR. ESRT is a hybrid model, which consists of a Lightweight CNN Backbone (LCB) and a Lightweight Transformer Backbone (LTB). Among them, LCB can dynamically adjust the size of the feature map to extract deep features with a low computational cost. LTB is composed of a series of Efficient Transformers (ET), which occupies a small GPU memory occupation, thanks to the specially designed Efficient Multi-Head Attention (EMHA). Extensive experiments show that ESRT achieves competitive results with low computational costs. Compared with the original Transformer which occupies 16,057M GPU memory, ESRT only occupies 4,191M GPU memory. All codes are available at https://github.com/luissen/ESRT.

Haiyue Zu, Jun Ge, Heting Xiao et al.

The reliance on large labeled datasets presents a significant challenge in medical image segmentation. Few-shot learning offers a potential solution, but existing methods often still require substantial training data. This paper proposes a novel approach that leverages the Segment Anything Model 2 (SAM2), a vision foundation model with strong video segmentation capabilities. We conceptualize 3D medical image volumes as video sequences, departing from the traditional slice-by-slice paradigm. Our core innovation is a support-query matching strategy: we perform extensive data augmentation on a single labeled support image and, for each frame in the query volume, algorithmically select the most analogous augmented support image. This selected image, along with its corresponding mask, is used as a mask prompt, driving SAM2's video segmentation. This approach entirely avoids model retraining or parameter updates. We demonstrate state-of-the-art performance on benchmark few-shot medical image segmentation datasets, achieving significant improvements in accuracy and annotation efficiency. This plug-and-play method offers a powerful and generalizable solution for 3D medical image segmentation.

Linlin Zhang, Xiang Yu, Abdulateef Daud et al.

Traffic cameras remain the primary source data for surveillance activities such as congestion and incident monitoring. To date, State agencies continue to rely on manual effort to extract data from networked cameras due to limitations of the current automatic vision systems including requirements for complex camera calibration and inability to generate high resolution data. This study implements a three-stage video analytics framework for extracting high-resolution traffic data such vehicle counts, speed, and acceleration from infrastructure-mounted CCTV cameras. The key components of the framework include object recognition, perspective transformation, and vehicle trajectory reconstruction for traffic data collection. First, a state-of-the-art vehicle recognition model is implemented to detect and classify vehicles. Next, to correct for camera distortion and reduce partial occlusion, an algorithm inspired by two-point linear perspective is utilized to extracts the region of interest (ROI) automatically, while a 2D homography technique transforms the CCTV view to bird's-eye view (BEV). Cameras are calibrated with a two-layer matrix system to enable the extraction of speed and acceleration by converting image coordinates to real-world measurements. Individual vehicle trajectories are constructed and compared in BEV using two time-space-feature-based object trackers, namely Motpy and BYTETrack. The results of the current study showed about +/- 4.5% error rate for directional traffic counts, less than 10% MSE for speed bias between camera estimates in comparison to estimates from probe data sources. Extracting high-resolution data from traffic cameras has several implications, ranging from improvements in traffic management and identify dangerous driving behavior, high-risk areas for accidents, and other safety concerns, enabling proactive measures to reduce accidents and fatalities.

Linlin Zhang, Xiang Yu, Armstrong Aboah et al.

Traffic volume data collection is a crucial aspect of transportation engineering and urban planning, as it provides vital insights into traffic patterns, congestion, and infrastructure efficiency. Traditional manual methods of traffic data collection are both time-consuming and costly. However, the emergence of modern technologies, particularly Light Detection and Ranging (LiDAR), has revolutionized the process by enabling efficient and accurate data collection. Despite the benefits of using LiDAR for traffic data collection, previous studies have identified two major limitations that have impeded its widespread adoption. These are the need for multiple LiDAR systems to obtain complete point cloud information of objects of interest, as well as the labor-intensive process of annotating 3D bounding boxes for object detection tasks. In response to these challenges, the current study proposes an innovative framework that alleviates the need for multiple LiDAR systems and simplifies the laborious 3D annotation process. To achieve this goal, the study employed a single LiDAR system, that aims at reducing the data acquisition cost and addressed its accompanying limitation of missing point cloud information by developing a Point Cloud Completion (PCC) framework to fill in missing point cloud information using point density. Furthermore, we also used zero-shot learning techniques to detect vehicles and pedestrians, as well as proposed a unique framework for extracting low to high features from the object of interest, such as height, acceleration, and speed. Using the 2D bounding box detection and extracted height information, this study is able to generate 3D bounding boxes automatically without human intervention.

Yutong Feng, Linlin Zhang, Hengyuan Cao et al.

Virtual Try-ON (VTON) is a practical and widely-applied task, for which most of existing works focus on clothes. This paper presents OmniTry, a unified framework that extends VTON beyond garment to encompass any wearable objects, e.g., jewelries and accessories, with mask-free setting for more practical application. When extending to various types of objects, data curation is challenging for obtaining paired images, i.e., the object image and the corresponding try-on result. To tackle this problem, we propose a two-staged pipeline: For the first stage, we leverage large-scale unpaired images, i.e., portraits with any wearable items, to train the model for mask-free localization. Specifically, we repurpose the inpainting model to automatically draw objects in suitable positions given an empty mask. For the second stage, the model is further fine-tuned with paired images to transfer the consistency of object appearance. We observed that the model after the first stage shows quick convergence even with few paired samples. OmniTry is evaluated on a comprehensive benchmark consisting of 12 common classes of wearable objects, with both in-shop and in-the-wild images. Experimental results suggest that OmniTry shows better performance on both object localization and ID-preservation compared with existing methods. The code, model weights, and evaluation benchmark of OmniTry will be made publicly available at https://omnitry.github.io/.

Linlin Zhang

Most existing document-level neural machine translation (NMT) models leverage a fixed number of the previous or all global source sentences to handle the context-independent problem in standard NMT. However, the translating of each source sentence benefits from various sizes of context, and inappropriate context may harm the translation performance. In this work, we introduce a data-adaptive method that enables the model to adopt the necessary and useful context. Specifically, we introduce a light predictor into two document-level translation models to select the explicit context. Experiments demonstrate the proposed approach can significantly improve the performance over the previous methods with a gain up to 1.99 BLEU points.

Junliang Guo, Zhirui Zhang, Linlin Zhang et al.

Adversarial attacks have shown the vulnerability of machine learning models, however, it is non-trivial to conduct textual adversarial attacks on natural language processing tasks due to the discreteness of data. Most previous approaches conduct attacks with the atomic \textit{replacement} operation, which usually leads to fixed-length adversarial examples and therefore limits the exploration on the decision space. In this paper, we propose variable-length textual adversarial attacks~(VL-Attack) and integrate three atomic operations, namely \textit{insertion}, \textit{deletion} and \textit{replacement}, into a unified framework, by introducing and manipulating a special \textit{blank} token while attacking. In this way, our approach is able to more comprehensively find adversarial examples around the decision boundary and effectively conduct adversarial attacks. Specifically, our method drops the accuracy of IMDB classification by $96\%$ with only editing $1.3\%$ tokens while attacking a pre-trained BERT model. In addition, fine-tuning the victim model with generated adversarial samples can improve the robustness of the model without hurting the performance, especially for length-sensitive models. On the task of non-autoregressive machine translation, our method can achieve $33.18$ BLEU score on IWSLT14 German-English translation, achieving an improvement of $1.47$ over the baseline model.