Infrared small and dim target detection is highly sensitive to background clutter because target responses are usually weak, sparse, and easily confused with non-target thermal structures. To improve detection robustness under complex infrared scenes while maintaining computational efficiency, this paper presents SASE-DETR, an improved end-to-end detector built on RT-DETR. The proposed framework enhances feature representation at three levels. First, a C2f-SASCG module is introduced into the backbone, where spatially adaptive sampling and channel gating are combined to strengthen local structural responses of small targets. Second, a Hierarchical Attention Fusion Block is designed to guide cross-scale feature interaction and reduce the propagation of background-dominated information. Third, a Distribution-aware Hierarchical Self-Attention encoder is incorporated to model feature response distributions during global semantic representation. Experiments on HIT-UAV and SIRST show that SASE-DETR improves [email protected] by 3.7% and 2.5% compared with RT-DETR-r18, respectively. Meanwhile, the proposed model contains 16.7M parameters and achieves an inference speed of approximately 84 FPS. Additional evaluation on IRSTD-1K further demonstrates its stable generalization under different infrared data distributions. These results indicate that SASE-DETR provides a favorable trade-off between detection accuracy, model complexity, and real-time inference for infrared small target detection.

Guoqiang Gong, Jiaxing Wang, Jin Xu, Deping Xiang, Zicheng Zhang, Leqi Shen, Yifeng Zhang, JunhuaShu JunhuaShu, ZhaolongXing ZhaolongXing, Zhen Chen, Pengzhang Liu, Ke Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The parameter-efficient adaptation of the image-text pre-training model CLIP for video-text retrieval is a prominent area of research. While CLIP is focused on image-level vision-language matching, video-text retrieval demands comprehensive understanding at the video level. Three key discrepancies emerge in the transfer from image-level to video-level: vision, language, and alignment. However, existing methods mainly focus on vision while neglecting language and alignment. In this paper, we propose Discrepancy Reduction in Vision, Language, and Alignment (DiscoVLA), which simultaneously mitigates all three discrepancies. Specifically, we introduce Image-Video Features Fusion to integrate image-level and video-level features, effectively tackling both vision and language discrepancies. Additionally, we generate pseudo image captions to learn fine-grained image-level alignment. To mitigate alignment discrepancies, we propose Image-To-Video Alignment Distillation, which leverages image-level alignment knowledge to enhance video-level alignment. Extensive experiments demonstrate the superiority of our DiscoVLA. In particular, on MSRVTT with CLIP (ViT-B/16), DiscoVLA outperforms previous methods by 2.2% R@1 and 7.5% R@sum. The code is available at https://github.com/LunarShen/DsicoVLA.

Video Large Language Models have demonstrated strong video understanding capabilities, yet their practical deployment is hindered by substantial inference costs caused by redundant video tokens. Existing pruning techniques fail to effectively exploit the spatiotemporal redundancy present in video data. To bridge this gap, we perform a systematic analysis of video redundancy from two perspectives: temporal context and visual context. Leveraging these insights, we propose Dynamic Density Pruning for Fast Video LLMs termed FastVID. Specifically, FastVID dynamically partitions videos into temporally ordered segments to preserve temporal structure and applies a density-based token pruning strategy to maintain essential spatial and temporal information. Our method significantly reduces computational overhead while maintaining temporal and visual integrity. Extensive evaluations show that FastVID achieves state-of-the-art performance across various short- and long-video benchmarks on leading Video LLMs, including LLaVA-OneVision, LLaVA-Video, Qwen2-VL, and Qwen2.5-VL. Notably, on LLaVA-OneVision-7B, FastVID effectively prunes $\textbf{90.3\%}$ of video tokens, reduces FLOPs to $\textbf{8.3\%}$, and accelerates the LLM prefill stage by $\textbf{7.1}\times$, while maintaining $\textbf{98.0\%}$ of the original accuracy. The code is available at https://github.com/LunarShen/FastVID.

The capabilities of large language models (LLMs) significantly depend on training data drawn from various domains. Optimizing domain-specific mixture ratios can be modeled as a bi-level optimization problem, which we simplify into a single-level penalized form and solve with twin networks: a proxy model trained on primary data and a dynamically updated reference model trained with additional data. Our proposed method, Twin Networks for bi-level DatA mixturE optiMization (TANDEM), measures the data efficacy through the difference between the twin models and up-weights domains that benefit more from the additional data. TANDEM provides theoretical guarantees and wider applicability, compared to prior approaches. Furthermore, our bi-level perspective suggests new settings to study domain reweighting such as data-restricted scenarios and supervised fine-tuning, where optimized mixture ratios significantly improve the performance. Extensive experiments validate TANDEM's effectiveness in all scenarios.

Low-Rank Adaptation (LoRA) offers a parameter-efficient paradigm for tuning large models. While recent spectral initialization methods improve convergence and performance over the naive “Noise \& Zeros” scheme, their extra computational and storage overhead undermines efficiency. In this paper, we establish update magnitude as the fundamental driver of LoRA performance and propose LoRAM, a magnitude-driven “Basis \& Basis” initialization scheme that matches spectral methods without their inefficiencies. Our key contributions are threefold: (i) Magnitude of weight updates determines convergence. We prove low-rank structures intrinsically bound update magnitudes, unifying hyperparameter tuning in learning rate, scaling factor, and initialization as mechanisms to optimize magnitude regulation. (ii) Spectral initialization succeeds via magnitude amplification. We demystify that the presumed knowledge-driven benefit of spectral component essentially arises from the boost in the weight update magnitude. (iii) A novel and compact initialization strategy, LoRAM, scales deterministic orthogonal bases using pretrained weight magnitudes to simulate spectral gains. Extensive experiments show that LoRAM serves as a strong baseline, retaining the full efficiency of LoRA while matching or outperforming spectral initialization across benchmarks.

Temporal sentence grounding in videos is a crucial task in vision-language learning. Its goal is retrieving a video segment from an untrimmed video that semantically corresponds to a natural language query. A video usually contains multiple semantic events, which are rarely isolated. They tend to be temporally ordered and semantically correlated (e.g., some event is often the precursor of another event). To precisely localize a semantic moment from a video, it is critical to effectively extract and aggregate multi-granularity contextual information, including the fine-grained local context around the moment-related video segment (in short snippet-level) and coarse-grained semantic correlation (in segment-level). Additionally, a second main insight in this work is that the above context aggregation should be favorably guided by the queries, rather than fully query-agnostic. Putting above ideas together, we here present a new network that does language-guided multi-granularity context aggregation. It is comprised of two major modules. The core of the first module is a novel language-guided temporal adaptive convolution (LTAC) devised to extract fine-grained information over video snippets around the ground-truth video segment. It decomposes a convolution into two channel-oriented / temporal-oriented ones. In particular, the convolutional channels are supposed to be more susceptible to queries, thus we learn to generate a dynamic channel-oriented kernel with respect to the querying sentence. As a second module, we propose a language-guided global relation block (LGRB) that extracts video-level context. It augments the contextual feature by using a multi-scale temporal attention that tackles the scale variation of ground-truth video segments, and a multi-modal semantic attention that relies on syntactic of the query. For the validation purpose, we have conducted comprehensive experiments on two popularly-adopted video benchmarks (i.e., ActivityNet Captions and Charades-STA). All experimental results and ablation studies have clearly corroborated the effectiveness of our model designs, outstripping prior state-of-the-art methods in terms of major performance metrics for the task.

Weakly-supervised temporal action localization aims to locate intervals of action instances with only video-level action labels for training. However, the localization results generated from video classification networks are often not accurate due to the lack of temporal boundary annotation of actions. Our motivating insight is that the temporal boundary of action should be stably predicted under various temporal transforms. This inspires a self-supervised equivariant transform consistency constraint. We design a set of temporal transform operations, including naive temporal down-sampling to learnable attention-piloted time warping. In our model, a localization network aims to perform well under all transforms, and another policy network is designed to choose a temporal transform at each iteration that adversarially brings localization result inconsistent with the localization network's. Additionally, we devise a self-refine module to enhance the completeness of action intervals harnessing temporal and semantic contexts. Experimental results on THUMOS14 and ActivityNet demonstrate that our model consistently outperforms the state-of-the-art weakly-supervised temporal action localization methods.

Abstract Speckle reduction is still an intractable task in ultrasound imaging field. Ultrasound speckle is usually described as multiplicative noise with its statistics following a Rayleigh or Gaussian distribution. To employ these two distributions effectively, the authors attempt to describe ultrasound speckle using a Weibull distribution, because it can include the Rayleigh distribution as a special case and also approximate a Gaussian distribution by varying its shape and scale parameters. The authors’ contribution in this paper is to propose a Weibull‐distribution‐based hybrid total variation (WHTV) method to reduce ultrasound speckle. The WHTV energy functional is convex and consists of a new data fidelity term and a new regularization term. The former is derived from the multiplicative Weibull model of ultrasound speckle based on the maximum likelihood criterion. The latter is a new edge‐weighted combination of the first‐ and second‐order total variation, with the advantage of preserving edges while alleviating the staircase effects. The minimization of the WHTV energy functional is implemented by the split Bregman algorithm. Experimental results on synthetic and real ultrasound images have demonstrated not only that the Weibull distribution is a better fitting model for the statistics of ultrasound speckle than other distributions such as Rayleigh, Gaussian, Gamma, and Nakagami, but also that the proposed WHTV method can achieve better despeckling performance than several state‐of‐the‐art variational methods.

Temporal action localization (TAL) in untrimmed videos recently receives tremendous research enthusiasm. To our best knowledge, this is the first attempt in the literature to explore this task under an unsupervised setting, hereafter referred to as action co-localization (ACL), where only the total count of unique actions that appear in the video set is known. To solve ACL, we propose a two-step ``clustering + localization" iterative procedure. The clustering step provides noisy pseudo-labels for the localization step, and the localization step provides temporal co-attention models that in turn improve the clustering performance. Using such two-step procedure, weakly-supervised TAL can be regarded as a direct extension of our ACL model. Technically, our contributions are two-folds: 1) temporal co-attention models, either class-specific or class-agnostic, learned from video-level labels or pseudo-labels in an iterative reinforced fashion; 2) new losses specially designed for ACL, including action-background separation loss and cluster-based triplet loss. Comprehensive evaluations are conducted on 20-action THUMOS14 and 100-action ActivityNet-1.2. On both benchmarks, the proposed model for ACL exhibits strong performances, even surprisingly comparable with state-of-the-art weakly-supervised methods. For example, previous best weakly-supervised model achieves 26.8% under [email protected] on THUMOS14, our new records are 30.1% (weakly-supervised) and 25.0% (unsupervised).

Video visual relation detection (VidVRD) aims to describe all interacting objects in a video. Different from relationships in static images, videos contain an addition temporal channel. A majority of existing works divide a video into short segments, predict relationships in each segment, and merge them. Such methods cannot capture relations involving long motions. Predicting the same relationship across neighboring video segments is also inefficient. To address these issues, this work proposes a novel sliding-window scheme to simultaneously predict short-term and long-term relationships. We run windows with different kernel sizes on object tracklets to generate sub-tracklet proposals with different duration, while the computational load is similar to that in segment-based methods. To fully utilize spatial and temporal information in videos, we construct one spatial and one temporal graph and employ Graph Convloutional Network to generate contextual embedding for tracklet proposal compatibility evaluation. We only predict relationships on highly-compatible proposal pairs. Our method achieves state-of-the-art performance on both ImageNet-VidVRD and VidOR dataset across multiple tasks. Especially for ImageNet-VidVRD, we obtain an average of 3% (R@50 from 8.07% to 11.21%) improvement under all evaluation metrics.

Temporal action localization is a recently-emerging task, aiming to localize video segments from untrimmed videos which contain specific actions. This work proposes a novel integrated temporal scale aggregation network (TSA-Net). Our main insight is that ensembling convolution filters with different dilation rates can effectively enlarge the receptive field with low computational cost, which inspires us to devise multi-dilation temporal convolution (MDC) block. Furthermore, to tackle video action instances with different durations, TSA-Net consists of multiple branches of sub-networks. Each of them adopts stacked MDC blocks with different dilation parameters, accomplishing a temporal receptive field specially optimized for specific-duration actions. We follow the formulation of boundary point detection, novelly detecting three kinds of critical points (i.e., starting / mid-point / ending) and pairing them for proposal generation. Comprehensive evaluations are conducted on THUMOS14. Our proposed TSA-Net demonstrates clear and consistent better performances and recalibrates new state-of-the-art on THUMOS14 benchmark.

Intensity inhomogeneity is one of the major obstacles for intensity‐based segmentation in many applications. The recently proposed kernel mapping (KM) method has exhibited excellent performance on segmenting various types of noisy images while it is not effective to handle intensity inhomogeneity. To overcome this drawback, this study presents a localised KM (LKM) method based on the fact that intensity inhomogeneity can be ignored in a local neighbourhood. The authors’ method first reconstructs the KM formulation of image segmentation in a neighbourhood of each pixel, and then such formulations for all pixels can be integrated together to derive the LKM energy functional. Minimisation of the energy functional is implemented by solving an equivalent convex‐relaxed problem whose optimisation can be quickly achieved via the split Bregman method. Experimental results on two‐phase segmentation and multiphase segmentation demonstrate competitive performance of the LKM method in the presence of intensity inhomogeneity and severe noise.