论文检索

Chuan Zhang 0003, Zihan Li, Zihao Xu, Xuhao Ren, Liehuang Zhu

Visual-audio Deepfake has become increasingly prevalent in today's online environment. Passive detection methods, lacking preventive measures, struggle with detecting unknown forgery techniques, limiting their effectiveness. While proactive detection methods offer greater robustness, unimodal watermarking approaches remain vulnerable in visual-audio Deepfake scenarios, posing challenges to reliable forensics. To address these challenges, we propose a novel Separable Visual-Audio waterMark framework, called SepVAMark, for proactive Deepfake detection. SepVAMark incorporates a multi-layer perceptron-based mixer layer to fuse intra-modality and inter-modality features from both audio and visual data. We introduce the concept of separable visual-audio watermark, along with a bimodal robust extractor for traceability and two unimodal semi-robust extractors for Deepfake detection. This design ensures reliable copyright protection for source audio-video content while enabling authenticity verification for redistributed content. Experimental results on the FakeAVCeleb dataset demonstrate that SepVAMark effectively detects a wide range of advanced Deepfake manipulations, outperforming existing single-modal and multi-modal watermarking methods with superior robustness.

Zhuo Su 0001, Jufeng Li, Yan Zhang 0002, Xin Li 0175, Fuwei Zhang, Yuxin Feng, Fan Zhou 0001

Existing learning-based dehazing methods perform well on synthetic data but struggle in real scenarios due to the domain gap, causing residual haze and detail loss. To address this, we propose a Multilevel Subspace Distribution Adapter (MSDA) to progressively reduce the feature distribution gap through hierarchical subspace modeling. We also introduce a Dual-Domain Synchronous Optimization (DDSO) strategy that jointly leverages synthetic supervision and adaptation to the real domain in a unified training scheme. Extensive experiments underscore the superiority of our approach and its excellence on no-reference image quality metrics.

Hanyuan Liu, Minshan Xie, Jinbo Xing, Chengze Li, Chi-Sing Leung, Tien-Tsin Wong

Video colorization is inherently challenging due to the need for accurate color inference and temporal consistency. In this paper, we present ColorDiffuser, an adaptation of a pre-trained text-to-image latent diffusion model for video colorization. By leveraging learned color priors from large-scale training, our method avoids costly retraining and enables controllable colorization via text prompts. To address the adaptation of an image model to video, we propose a novel Short- and Long-distance Cross-Frame Attention (SL-CFA) module combined with an amortized sampling strategy to unify the color latent over time. By incorporating information from nearby and distant frames, the model achieves better consistency for long video sequences, even with problematic disocclusion. To mitigate visual detail loss and color bleeding from compressed latent representations, we introduce a video colorization VAE model that incorporates semantic boundaries and grayscale inputs. Extensive experiments on benchmark datasets demonstrate that ColorDiffuser achieves state-of-the-art performance in color fidelity, temporal consistency, and visual quality, while offering diverse and controllable outputs. Our project page can be accessed at: https://colordiffuser.github.io.

Le Liu 0008, Shizhou Zhang, Di Xu 0010

Effectively communicating uncertainty in ensemble hurricane forecasts poses a significant multimedia challenge, requiring the integration of spatial, temporal, and perceptual dimensions. We introduce SUVIS, a stereoscopic visualization system that encodes forecast ensembles into an immersive, layered media experience. SUVIS transforms multidimensional ensemble data into animated stereoscopic representations, mapping time to vertical depth, intensity to texture color, and forward speed to motion flow, while semi-transparent glyphs represent evolving impact areas. A progressive sampling strategy ensures spatial clarity across depth layers. Rendered on a glasses-free stereoscopic display, SUVIS frames uncertainty visualization as a media encoding problem, synthesizing motion, depth, and spatial abstraction to align with human perception. A user study with 51 participants demonstrates that SUVIS supports high accuracy in spatial tasks and enables interpretation of dynamic storm attributes. These results highlight the system's potential to advance perceptual uncertainty communication through multimedia representation and immersive visual encoding.

Yan Rong, Jinting Wang, Guangzhi Lei, Shan Yang 0001, Li Liu 0036

Multimodality-to-Multiaudio (MM2MA) generation faces significant challenges in synthesizing diverse and contextually aligned audio types (e.g., sound effects, speech, music, and songs) from multimodal inputs (e.g., video, text, images), owing to the scarcity of high-quality paired datasets and the lack of robust multi-task learning frameworks. Recently, multi-agent system shows great potential in tackling the above issues. However, directly applying it to MM2MA task presents three critical challenges: (1) inadequate fine-grained understanding of multimodal inputs (especially for video), (2) the inability of single models to handle diverse audio events, and (3) the absence of self-correction mechanisms for reliable outputs. To this end, we propose AudioGenie, a novel training-free multi-agent system featuring a dual-layer architecture with a generation team and a supervisor team. For the generation team, a fine-grained task decomposition and an adaptive Mixture-of-Experts (MoE) collaborative entity are designed for detailed comprehensive multimodal understanding and dynamic model selection, and a trial-and-error iterative refinement module is designed for self-correction. The supervisor team ensures temporal-spatial consistency and verifies outputs through feedback loops. Moreover, we build MA-Bench, the first benchmark for MM2MA tasks, comprising 198 annotated videos with multi-type audios. Experiments demonstrate that our AudioGenie achieves state-of-the-art (SOTA) or comparable performance across 9 metrics in 8 tasks. User study further validates the effectiveness of our method in terms of quality, accuracy, alignment, and aesthetic. The project website with audio samples can be found at https://audiogenie.github.io/.

Dongjian Yu, Weiqing Min, Xin Jin 0005, Qian Jiang, Shuqiang Jiang

Food nutrition assessment plays a crucial role in maintaining health, preventing diseases, and promoting scientific dietary habits. However, existing nutrition assessment methods often fail to fully consider the relationships between tasks, leading to limited overall performance. Specifically, these methods suffer from three major challenges: (1) task conflicts, where different tasks compete during joint optimization, leading to suboptimal overall performance; (2) varying training difficulties among tasks, leading to imbalanced learning and subpar model generalization; and (3) the small-scale and complex distribution of datasets, which limits the robustness of learned representations. To address these issues, we propose a novel method that reduces interference between tasks, dynamically focuses on more challenging tasks, and incorporates 3D spatial awareness to enhance multi-modal feature representation. First, we decouple the prediction network from the backbone and introduce a CAMTH (Cross-Attention-Based Multi-Task Head Module), effectively mitigating task interference and fully leveraging each task's learning potential. Second, we improve the loss function to adaptively focus on more challenging tasks, improving overall model performance. Third, we design a 3D-FEM (3D Feature Extraction Module) and MMFF (Multi-Modal Feature Fusion Module), enabling the model to fully exploit the spatial information of food and enhance the food's multi-modal feature representation. We validate our method through extensive experiments on the Nutrition5K dataset, comparing it with state-of-the-art (SOTA) models. The results show that our method achieves superior performance in nutrition estimation, demonstrating the effectiveness of our method.

Shuai Yu 0002, Xiaoliang He, Kangjie Dong, Yi Yu 0001

Semi-supervised singing melody extraction (SSME) is one of the key tasks in the field of music information retrieval (MIR). However, there are two critical issues that remain to be addressed in data limited scenarios. Firstly, the prior unsupervised domain adaptation methods for SSME typically rely on learning domain-agnostic features at holistic level, which ignores the associations between holistic information (i.e., fundamental frequency) and fine-grained information (i.e., tone and octave). Secondly, the fine-grained information can be utilized to judge the availability of unlabeled data, which is ignored by prior methods. There is a lack of a consistency regularization method that utilizes fine-grained information to validate the availability of unlabeled data. To address these issues, in this paper, we propose a novel two-stage decoupling unsupervised domain adaptation framework for semi-supervised singing melody extraction, termed as DUDA. Specifically, in the first stage, we decouple the holistic information into fine-grained information: tone and octave, and narrow the domain gap at the tone and octave level, respectively. This enables the model to align the tone-octave information between source and target domains for better feature distribution. Then, we leverage the learned domain-agnostic fine-grained features as additional information to obtain domain-agnostic holistic features. We also suggest to align intra-domain, inter-domain, and sample-level features to further improve the performances. In the second stage, we propose a novel tone-octave consistency regularization method by leveraging the extracted fine-grained information to judge the availability of unlabeled data. We evaluate our proposed framework on several well-known public datasets, and the conducted experiments demonstrate the effectiveness of our method.

Yue Hou, Yingke Su, Junran Wu, Ke Xu 0001

Out-of-distribution (OOD) detection for graph-structured data remains a challenging problem, particularly when test-time OOD samples deviate significantly from the training outliers. Existing methods are typically optimized to capture the features within the in-distribution (ID) training data, but often fail to model the transitional region near the boundary between ID and OOD samples. Moreover, since data distributions are usually governed by multiple latent factors, pre-trained models constrained by the scope and diversity of training data struggle to represent the full spectrum of sample characteristics and distributional boundaries. To address this dilemma, we propose a novel test-time graph OOD detection method, termed D2GO, that constructs and dynamically updates ID and OOD graphon dictionaries for OOD score calibration, without requiring fine-tuning. Specifically, D2GO estimates graphons from test graphs and employs a mix-up strategy to generate boundary samples, eliminating the need for exposing auxiliary datasets or training graphs. Priority queues are utilized to expand the ID and OOD dictionaries by incorporating diverse graphons based on pseudo-labels at test-time, and the OOD scores are calibrated by computing the similarity between test samples and both graphon dictionaries. Extensive experiments on real-world datasets show that D2GO significantly outperforms existing state-of-the-art methods in OOD detection.

Ting Li, Songtao Li, Shuaifeng Li, Xiaolin Qin, Maoyuan Zhao, Luping Ji, Mao Ye 0001

Visible-infrared object detection has gained significant attention because of its applications in autonomous driving, video surveillance, and related fields. The effective fusion of multimodal information is fundamental to its success. The existing approaches concentrate on improving the pixel-level fusion mechanisms; detection performance has reached a plateau. We propose a new framework for SAM-guided semantic knowledge fusion ( SemFusion ). The core idea is to leverage semantic priors from large models while incorporating a lightweight cross-modal fusion strategy. Specifically, our method comprises two stages. In the first stage, the Flow-Guided RGB Feature Alignment (FGRA) module establishes object-aware correspondences between multimodalities based on SAM-generated masks. This ensures semantic-level feature matching by deformable convolution alignment. In the second stage, the Semantic Knowledge Distillation (SKD) strategy facilitates the transfer of large-model knowledge to the detection model through SAM feature, offset, and mask level distillations. For the detector model, three blocks are designed to augment any off-the-shelf detector. They are deformable cross-modal alignment, spatio-channel preliminary fusion, and mask-guided feature refinement. By alignment with SAM masks, semantic alignment and fusion can be achieved, breaking the pixel-level fusion barrier. Extensive experiments demonstrate that our method, as a plugin, exhibits superior performance on the DroneVehicle, VEDAI, and LLVIP datasets. Code is available at https://github.com/liting1018/SemFusion.

Woo Yi Yang, Jiarui Wang, Sijing Wu, Huiyu Duan, Yuxin Zhu, Liu Yang, Kang Fu, Guangtao Zhai, Xiongkuo Min

The rapid advancement in generative artificial intelligence have enabled the creation of 3D human faces (HFs) for applications including media production, virtual reality, security, healthcare, and game development, etc. However, assessing the quality and realism of these AI-generated 3D human faces remains a significant challenge due to the subjective nature of human perception and innate perceptual sensitivity to facial features. To this end, we conduct a comprehensive study on the quality assessment of AI-generated 3D human faces. We first introduce Gen3DHF, a large-scale benchmark comprising 2,000 videos of AI-Generated 3D Human Faces along with 4,000 Mean Opinion Scores (MOS) collected across two dimensions, i.e., quality and authenticity, 2,000 distortion-aware saliency maps and distortion descriptions. Based on Gen3DHF, we propose LMME3DHF, a Large Multimodal Model (LMM)-based metric for Evaluating 3DHF capable of quality and authenticity score prediction, distortion-aware visual question answering, and distortion-aware saliency prediction. Experimental results show that LMME3DHF achieves state-of-the-art performance, surpassing existing methods in both accurately predicting quality scores for AI-generated 3D human faces and effectively identifying distortion-aware salient regions and distortion types, while maintaining strong alignment with human perceptual judgments. Both the Gen3DHF database and the LMME3DHF will be released upon the publication.

Hantao Zhou, Rui Yang 0041, Longxiang Tang, Guanyi Qin, Runze Hu, Xiu Li 0001

Image assessment aims to evaluate the quality and aesthetics of images and has been applied across various scenarios, such as natural and AIGC scenes. Existing methods mostly address these sub-tasks or scenes individually. While some works attempt to develop unified image assessment models, they have struggled to achieve satisfactory performance or cover a broad spectrum of assessment scenarios. In this paper, we present Gamma, a Generic imAge assessMent model using Mixture of Assessment Experts, which can effectively assess images from diverse scenes through mixed-dataset training. Achieving unified training in image assessment presents significant challenges due to annotation biases across different datasets. To address this issue, we first propose a Mixture of Assessment Experts (MoAE) module, which employs shared and adaptive experts to dynamically learn common and specific knowledge for different datasets, respectively. In addition, we introduce a Scene-based Differential Prompt (SDP) strategy, which uses scene-specific prompts to provide prior knowledge and guidance during the learning process, further boosting adaptation for various scenes. Our Gamma model is trained and evaluated on 12 datasets spanning 6 image assessment scenarios. Extensive experiments show that our unified Gamma outperforms other state-of-the-art mixed-training methods by significant margins while covering more scenes. Codes are available at https://github.com/zht8506/Gamma.

Yize Song, Yunqing Chen, Zhou Wang, Cheng Chen 0024, Ruoxiu Xiao

The pelvis is a high-incidence region for trauma, and its precise segmentation is vital for clinical diagnosis and treatment. Although deep learning has achieved high precision for pelvis segmentation, it is still confronted with challenges such as the scarcity of medical data and the high cost of annotation. Recently, generative models have offered a viable solution through synthetic data augmentation. Thus, we propose a novel text-conditioned generative framework that simultaneously produces high-fidelity CT images and their corresponding segmentation masks, with a special focus on accurately generating symmetric pelvic structures, including the left hip bone, right hip bone and sacrum. Firstly, we fine-tuned a medical text encoder to transform detailed descriptions into precise generation conditions. Considering the easily overlooked symmetry attribute in pelvic bone images, we introduced a novel coordinate-aware enhancement module that incorporates bone-specific centroid coordinates for symmetrical awareness. Finally, we expanded it to multi-task learning for generation of paired pelvic images and segmentation labels. To test our method, we released a pelvic CT image dataset with textual description (CT-PelvisText) and then transferred it to downstream segmentation using the generative pelvic image. Our experiments demonstrate the reliability of our method, which contributes to accurate pelvic segmentation. This work can also be easily extended to other medical images with symmetry properties, which provides potential for efficient learning in small-sample datasets. Our code and dataset are available at: https://github.com/CurellaSong/TSA_LDM.

Zeyu Huang, Juyuan Wang, Longfeng Chen, Boyi Xiao, Leng Cai, Yawen Zeng, Jin Xu 0014

Given the significant advances in Large Vision Language Models (LVLMs) in reasoning and visual understanding, mobile agents are rapidly emerging to meet users' automation needs. However, existing evaluation benchmarks are disconnected from the real world and fail to adequately address the diverse and complex requirements of users. From our extensive collection of user questionnaire, we identified five tasks: Multi-App, Vague, Interactive, Single-App, and Unethical Instructions. Around these tasks, we present MVISU-Bench, a bilingual benchmark that includes 404 tasks across 137 mobile applications. Furthermore, we propose Aider, a plug-and-play module that acts as a dynamic prompt prompter to mitigate risks and clarify user intent for mobile agents. Our Aider is easy to integrate into several frameworks and has successfully improved overall success rates by 19.55% compared to the current state-of-the-art (SOTA) on MVISU-Bench. Specifically, it achieves success rate improvements of 53.52% and 29.41% for unethical and interactive instructions, respectively. Through extensive experiments and analysis, we highlight the gap between existing mobile agents and real-world user expectations.

Qin Li 0010, Congcong Xiao, Limei Liu, Han Peng, Junfeng Yang

Skeleton-based human action recognition (HAR) is greatly affected by abnormal situations in real-world scenarios, like occlusions and performance limitations of motion capture devices. Although recent research has enhanced the robustness of recognition by incorporating occlusion simulation in model training, it is still insufficient to effectively handle the complex and diverse abnormal situations in real-world scenarios. To address this issue, we propose SCCEAP, a novel framework combining fine-grained Skeleton Compression and Complementary Enhanced Adaptive feature fusion with the supervision of branch-stage text Prompts, for robust skeleton-based HAR. Our contributions lie in three aspects. First, the fine-grained skeleton compression is designed to generate multi-granularity skeleton sequences with diverse spatial details by fusing joints in the human skeleton according to their joint reliabilities and correlations. Then, we devise the complementary enhanced adaptive feature fusion, which utilizes motion details and stable semantic descriptions of motion features of the uncompressed and compressed skeleton sequences respectively, for complementary enhancement and adaptive feature fusion. Third, the branch-stage composite text-prompt supervision is performed to integrate both branch-wise and stage-wise text-prompt supervision for improving the ability to learn fine-grained spatiotemporal relationships of motion features. Experiments on three benchmark datasets-NTU RGB+D, NTU RGB+D 120, and Kinetics-400-demonstrate that SCCEAP achieves the state-of-the-art (SOTA) results, excelling on both normal and noisy skeleton data.

Kaixin Li, Ziyang Meng, Hongzhan Lin 0001, Ziyang Luo, Yuchen Tian, Jing Ma 0004, Zhiyong Huang 0010, Tat-Seng Chua

Recent advancements in Multi-modal Large Language Models (MLLMs) have led to significant progress in developing GUI agents for general tasks such as web browsing and mobile phone use. However, their application in professional domains remains under-explored. These specialized workflows introduce unique challenges for GUI perception models, including high-resolution displays and complex environments which lead to smaller target sizes. In this paper, we introduce ScreenSpot-Pro, a new benchmark designed to rigorously evaluate the grounding capabilities of MLLMs in high-resolution professional settings. The benchmark comprises authentic high-resolution images from a variety of professional domains with expert annotations. It spans 23 applications across five industries and three operating systems. Existing GUI grounding models perform poorly on this dataset, with the best model achieving only 18.9%. Our experiments reveal that strategically reducing the search area enhances accuracy. Based on this insight, we propose ScreenSeekeR, a visual search method that utilizes the GUI knowledge of a strong planner to guide a cascaded search, achieving state-of-the-art performance with 48.1% without any additional training. We hope that our benchmark and findings will advance the development of GUI agents for professional settings.

Jiawei Zhang 0011, Xiaoli Jiang, Hao Wang 0060, Lin Yuan 0002, Xiangyang Luo 0001, Bin Ma 0003, Jinwei Wang

With the rapid development of the diffusion models, numerous exquisitely generated images have significantly increased the risk of image misuse and abuse. Despite various AI parties and companies having devoted themselves to embedding watermarks into the generated images to curb the potential detriments, the isolated embedding from the generation process makes the watermarks vulnerable to watermark removal networks. To address this issue, we propose a novel generative image watermark scheme, dubbed Diffusion Visible Watermark (DVW), which can generate watermarked images in one step without additional training or fine-tuning of the diffusion models. Specifically, DVW introduces a masked distribution alignment strategy to fuse the watermark distribution with a Gaussian noise distribution. By iterative denoising the fused aligned distribution with the pretraining diffusion models, the watermarked images with coordinated and unified distribution can be generated with natural robustness against removal. In addition, we design and integrate a dynamic transparency module to adaptively control the watermark coverage degree for better visual quality. Comprehensive experiments and analysis are conducted on two representative kinds of diffusion models, GLIDE and StableDiffusion, to prove the superior and generic robustness of our DVW against watermark removal without sacrificing the generation ability of the diffusion models.

Yunyu Zou, Yishu Liu 0001, Jun Liang 0002, Bingzhi Chen

Cross-Domain Few-Shot Learning (CD-FSL) aims to transfer knowledge acquired from a source domain with abundant data to the target domain with limited labeled samples. Recent advancements have enhanced model generalization through Perturbation Augmentation (PA), facilitating more effective knowledge transfer. However, PA-based CD-FSL methods still suffer from two critical challenges, i.e., (1) limited diversity of augmented samples, making it difficult to cover the true distribution of unseen domains, and (2) conflicting gradients during model optimization, where augmented and original samples drive the model's optimization in opposing directions. To address these issues, we propose a novel PA-based framework with Style-Decoupled Augmentation (SDA) and Gradient-Conflict Adjustment (GCA) for Cross-Domain Few-Shot Learning, which is termed ''SG-FSL''. Specifically, SDA decouples the source domain style into style weights and basis styles, generating diverse unseen styles by perturbing the style weights to reweight the basis styles. Meanwhile, GCA leverages the angular relationships between the domain-specific gradient directions of augmented and original features, adaptively adjusting the gradient directions of original features to ensure that the model acquires diverse domain knowledge without interference, guiding it toward conflict-free optimization. Comprehensive experiments on multiple benchmark datasets consistently demonstrate the effectiveness and superiority of our method over state-of-the-art baselines.

Ting Xiao 0002, Minqian Sun, Yiqing Xia, Zhe Wang 0002

Existing prototype learning-based Multiple Instance Learning (MIL) methods mainly focus on learning a single set of prototypes for each class or generating a generic prototype from the overall data distribution. This design forces the model to compress general and heterogeneous features into identical prototype embeddings, prioritizing general features over subtle but discriminative features. Additionally, these methods often guide prototype updates by jointly optimizing attention score distributions and the distances between instances and prototypes, resulting in prototype biases due to over-concentration. To address these issues, we propose a dual prototype learning MIL (DP-MIL) framework that introduces two distinct sets of prototypes: primary prototypes, which capture general WSI features, and boundary prototypes, which capture discriminative features near the decision boundary. The DP-MIL framework employs three prototype-tailored losses: an alienation loss to encourage primary prototypes to be distant from decision boundaries, an affinity loss to anchor boundary prototypes near these boundaries, and a distance loss to enforce separation between the two prototype sets. To mitigate prototype semantic drift during training, we introduce a prototype joint updating and refinement strategy: for each prototype, we use its corresponding global token to filter out the most similar instances to momentum update the corresponding prototype set, while the boundary prototype set is refined with the mean pooled feature of hard samples. Extensive experiments on four datasets demonstrate the effectiveness of our DP-MIL framework and prototype updating strategy.

Haoyu Shi, Huaiwen Zhang

Text-to-Motion Retrieval (TMR) is a challenging task to retrieve relevant motion sequences with the natural language description. Existing TMR methods primarily utilize single embeddings to represent and align text and motion sequences. However, real-world motion sequences typically contain multiple sequential actions with intricate semantics, which are hard to precisely capture by single embedding. Additionally, relying solely on naive contrastive training to capture high-level semantics may struggle to perceive and capture fine-grained action details necessary for precise text-motion alignment. In this work, we propose a novel Sequence-Event Semantic Consistent Learning (SECL) framework for 3D human motion retrieval. Specifically, we introduce a self-supervised learning strategy to incorporate fine-grained action details into the motion representations via the generative feedback from the diffusion model. We design a parameter-free sequence-level interaction to explore coarse-grained alignment and an event-level interaction that utilizes several learnable queries to capture event semantics in a shared learning manner for fine-grained alignment. Furthermore, an inter-consistency loss is introduced to align the event semantics between the motion and corresponding text, and an intra-diversity loss is designed to encourage event features to attend to different contents, effectively capturing the rich action information. Finally, we modify the traditional contrastive alignment objective and propose an importance-sampling strategy to emphasize harder negatives for discriminative representation learning. Extensive experiments show that our method significantly outperforms existing methods in text-to-motion retrieval and other challenging tasks, e.g., human interaction recognition and motion temporal localization.

Yusen Wang 0002, Huan Zhou, Yu Jiang 0007, Chunxia Xiao

We propose GSAPro, a Gaussian Splatting based 3D surface reconstruction framework that exhibits robustness across diverse scales of scenes. Previous research has leveraged photometric consistency constraints or prior information as guidance to enhance the reconstruction accuracy. However, error estimation and noise inevitably exist in these priors. Applying a strict geometric filter removes a large amount of reliable information, resulting in a deterioration of the quality of guided reconstruction. Regarding possible errors in the initial guidance, GSAPro can continuously improve the accuracy of the guidance through a joint optimization strategy. The Gaussian Branch integrates reliable geometric and color constraints, thus providing more accurate geometric parameters for the Prior Branch compared to its current state guidance parameters. The Prior Branch, through photometric selection and propagation, obtains more accurate geometric parameters from the state geometric parameters and rendered parameters. Then GSAPro uses these parameters to guide the optimization of the Gaussian Branch. Regarding the problem of noise existing in the guidance, we train the Semantic Aware Module to predict the noise by utilizing the image information, thus improving the accuracy. Moreover, we also introduce a Distillation Module to mitigate the excessive splitting of Gaussians that is caused by the implementation of additional constraints. Experiments demonstrate that our method exhibits SOTA performance and has stronger robustness against scenes of different scales.