Yuyu Guo

Xinyu Lyu, Lianli Gao, Yuyu Guo et al.

The performance of current Scene Graph Generation models is severely hampered by some hard-to-distinguish predicates, e.g., "woman-on/standing on/walking on-beach" or "woman-near/looking at/in front of-child". While general SGG models are prone to predict head predicates and existing re-balancing strategies prefer tail categories, none of them can appropriately handle these hard-to-distinguish predicates. To tackle this issue, inspired by fine-grained image classification, which focuses on differentiating among hard-to-distinguish object classes, we propose a method named Fine-Grained Predicates Learning (FGPL) which aims at differentiating among hard-to-distinguish predicates for Scene Graph Generation task. Specifically, we first introduce a Predicate Lattice that helps SGG models to figure out fine-grained predicate pairs. Then, utilizing the Predicate Lattice, we propose a Category Discriminating Loss and an Entity Discriminating Loss, which both contribute to distinguishing fine-grained predicates while maintaining learned discriminatory power over recognizable ones. The proposed model-agnostic strategy significantly boosts the performances of three benchmark models (Transformer, VCTree, and Motif) by 22.8\%, 24.1\% and 21.7\% of Mean Recall (mR@100) on the Predicate Classification sub-task, respectively. Our model also outperforms state-of-the-art methods by a large margin (i.e., 6.1\%, 4.6\%, and 3.2\% of Mean Recall (mR@100)) on the Visual Genome dataset.

Chaofan Zheng, Xinyu Lyu, Yuyu Guo et al.

Scene Graph Generation (SGG) represents objects and their interactions with a graph structure. Recently, many works are devoted to solving the imbalanced problem in SGG. However, underestimating the head predicates in the whole training process, they wreck the features of head predicates that provide general features for tail ones. Besides, assigning excessive attention to the tail predicates leads to semantic deviation. Based on this, we propose a novel SGG framework, learning to generate scene graphs from Head to Tail (SGG-HT), containing Curriculum Re-weight Mechanism (CRM) and Semantic Context Module (SCM). CRM learns head/easy samples firstly for robust features of head predicates and then gradually focuses on tail/hard ones. SCM is proposed to relieve semantic deviation by ensuring the semantic consistency between the generated scene graph and the ground truth in global and local representations. Experiments show that SGG-HT significantly alleviates the biased problem and chieves state-of-the-art performances on Visual Genome.

Lianli Gao, Xinyu Lyu, Yuyu Guo et al.

Scene graph generation aims to detect visual relationship triplets, (subject, predicate, object). Due to biases in data, current models tend to predict common predicates, e.g. "on" and "at", instead of informative ones, e.g. "standing on" and "looking at". This tendency results in the loss of precise information and overall performance. If a model only uses "stone on road" rather than "stone blocking road" to describe an image, it may be a grave misunderstanding. We argue that this phenomenon is caused by two imbalances: semantic space level imbalance and training sample level imbalance. For this problem, we propose DB-SGG, an effective framework based on debiasing but not the conventional distribution fitting. It integrates two components: Semantic Debiasing (SD) and Balanced Predicate Learning (BPL), for these imbalances. SD utilizes a confusion matrix and a bipartite graph to construct predicate relationships. BPL adopts a random undersampling strategy and an ambiguity removing strategy to focus on informative predicates. Benefiting from the model-agnostic process, our method can be easily applied to SGG models and outperforms Transformer by 136.3%, 119.5%, and 122.6% on mR@20 at three SGG sub-tasks on the SGG-VG dataset. Our method is further verified on another complex SGG dataset (SGG-GQA) and two downstream tasks (sentence-to-graph retrieval and image captioning).

Cheng Chen, Pengpeng Zeng, Yuyu Guo et al.

Low-Rank Adaptation (LoRA) has emerged as a promising paradigm for Continual Learning. It independently updates its low-rank factors ($A$ and $B$), creating a composite update to the full weight matrix through their interaction. To prevent catastrophic forgetting, this update should remain orthogonal to the task-specific subspace that contains previously learned knowledge. However, we identify that this composite update systematically violates this orthogonality, reintroducing interference and undermining stability. Furthermore, naively enforcing this orthogonality compromises plasticity, disrupting the delicate stability-plasticity trade-off. To resolve these issues, we propose \textbf{Janus-LoRA}, a framework that restores this balance through two novel components. Specifically, we first introduce Gradient Rectification, a closed-form solution that mathematically decouples LoRA's factor updates, enforcing orthogonality against the historical knowledge subspace identified by an efficient Online Estimation. Next, to enhance plasticity, we introduce a Decoupled Margin Loss that promotes feature-level separation by pushing new feature representations away from old ones, thus creating distinct, low-interference regions for new learning. Comprehensive experiments on challenging benchmarks demonstrate that by harmonizing parameter-level orthogonality with feature-level separation, Janus-LoRA achieves a superior balance and establishes new state-of-the-art performance.

Xinyu Lyu, Jingwei Liu, Yuyu Guo et al.

The task of dynamic scene graph generation (DynSGG) aims to generate scene graphs for given videos, which involves modeling the spatial-temporal information in the video. However, due to the long-tailed distribution of samples in the dataset, previous DynSGG models fail to predict the tail predicates. We argue that this phenomenon is due to previous methods that only pay attention to the local spatial-temporal information and neglect the consistency of multiple frames. To solve this problem, we propose a novel DynSGG model based on multi-task learning, DynSGG-MTL, which introduces the local interaction information and global human-action interaction information. The interaction between objects and frame features makes the model more fully understand the visual context of the single image. Long-temporal human actions supervise the model to generate multiple scene graphs that conform to the global constraints and avoid the model being unable to learn the tail predicates. Extensive experiments on Action Genome dataset demonstrate the efficacy of our proposed framework, which not only improves the dynamic scene graph generation but also alleviates the long-tail problem.

Yuyu Guo, Wenjie Yang, Siyuan Yang et al.

To fulfill user instructions, autonomous web agents must contend with the inherent complexity and volatile nature of real-world websites. Conventional paradigms predominantly rely on Supervised Fine-Tuning (SFT) or Offline Reinforcement Learning (RL) using static datasets. However, these methods suffer from severe distributional shifts, as offline trajectories fail to capture the stochastic state transitions and real-time feedback of unconstrained wide web environments. In this paper, we propose a robust Online Reinforcement Learning WebAgent, designed to optimize its policy through direct, iterative interactions with unconstrained wide websites. Our approach comprises three core innovations: 1) Hierarchical Multi-Task Fine-tuning: We curate a comprehensive mixture of datasets categorized by functional primitives -- Planning, Acting, and Grounding -- establishing a Vision-Language Model (VLM) with strong instruction-following capabilities for Web GUI tasks. 2) Online Agentic RL in the Wild: We develop an online interaction environment and fine-tune the VLM using a specialized RL pipeline. We introduce a Hybrid Reward Mechanism that combines a ground-truth-agnostic WebJudge for holistic outcome assessment with a Rule-based Decision Tree (RDT) for progress reward. This system effectively mitigates the credit assignment challenge in long-horizon navigation. Notably, our RL-enhanced model achieves a 38.1\% success rate (pass@5) on WebArena, outperforming all existing monolithic baselines. 3) Operator Agent: We introduce a modular agentic framework, namely \textbf{OpAgent}, orchestrating a Planner, Grounder, Reflector, and Summarizer. This synergy enables robust error recovery and self-correction, elevating the agent's performance to a new State-of-the-Art (SOTA) success rate of \textbf{71.6\%}.

Cheng Chen, Yuyu Guo, Pengpeng Zeng et al.

Vision-Language Models (VLMs) create a severe visual feature bottleneck by using a crude, asymmetric connection that links only the output of the vision encoder to the input of the large language model (LLM). This static architecture fundamentally limits the ability of LLMs to achieve comprehensive alignment with hierarchical visual knowledge, compromising their capacity to accurately integrate local details with global semantics into coherent reasoning. To resolve this, we introduce Cross-Layer Injection (CLI), a novel and lightweight framework that forges a dynamic many-to-many bridge between the two modalities. CLI consists of two synergistic, parameter-efficient components: an Adaptive Multi-Projection (AMP) module that harmonizes features from diverse vision layers, and an Adaptive Gating Fusion (AGF) mechanism that empowers the LLM to selectively inject the most relevant visual information based on its real-time decoding context. We validate the effectiveness and versatility of CLI by integrating it into LLaVA-OneVision and LLaVA-1.5. Extensive experiments on 18 diverse benchmarks demonstrate significant performance improvements, establishing CLI as a scalable paradigm that unlocks deeper multimodal understanding by granting LLMs on-demand access to the full visual hierarchy.

Wenjie Yang, Hang Yu, Yuyu Guo et al.

In this work, we address the critical yet underexplored challenge of symmetric multimodal-to-multimodal (MM2MM) retrieval, where queries and contexts are interchangeable. Existing universal multimodal retrieval works struggle with this task, as they are constrained by the labeled asymmetric datasets used. We produce SOLAR (Self-supervised jOint LeArning for symmetric multimodal Retrieval), a novel two-stage self-supervised framework that leverages readily available unlabeled web-scale image-text pairs. Based on the observation that both semantic alignment and discrepancies exist between two modalities, in the first stage, we learn the intersection mask of image-text pair, allowing us to align intersection while preserving semantic of difference. In the second stage, the learned mask is further utilized to construct positive and hardnegative samples via masking different parts of image/text, which enable us to conduct self-supervised multimodal embedding learning. Complementing this framework, we present a new benchmark featuring high-quality human-verified positive and hard-negative pairs to evaluate symmetric MM2MM retrieval under realistic conditions, as well as the corresponding pipeline. Extensive experiments against ten SOTA methods show SOLAR surpasses the strongest supervised VLM by 7.08 points on this benchmark, with over 50x fewer model parameters and a 5x smaller embedding dimension. Code and benchmark will be available soon.

Yuyu Guo, Lei Bi, Zhengbin Zhu et al.

Automated segmentation of left ventricular cavity (LVC) in temporal cardiac image sequences (multiple time points) is a fundamental requirement for quantitative analysis of its structural and functional changes. Deep learning based methods for the segmentation of LVC are the state of the art; however, these methods are generally formulated to work on single time points, and fails to exploit the complementary information from the temporal image sequences that can aid in segmentation accuracy and consistency among the images across the time points. Furthermore, these segmentation methods perform poorly in segmenting the end-systole (ES) phase images, where the left ventricle deforms to the smallest irregular shape, and the boundary between the blood chamber and myocardium becomes inconspicuous. To overcome these limitations, we propose a new method to automatically segment temporal cardiac images where we introduce a spatial sequential (SS) network to learn the deformation and motion characteristics of the LVC in an unsupervised manner; these characteristics were then integrated with sequential context information derived from bi-directional learning (BL) where both chronological and reverse-chronological directions of the image sequence were used. Our experimental results on a cardiac computed tomography (CT) dataset demonstrated that our spatial-sequential network with bi-directional learning (SS-BL) method outperformed existing methods for LVC segmentation. Our method was also applied to MRI cardiac dataset and the results demonstrated the generalizability of our method.

Yuyu Guo, Jingqiu Zhang, Lianli Gao

Automatically describing videos with natural language is a fundamental challenge for computer vision and natural language processing. Recently, progress in this problem has been achieved through two steps: 1) employing 2-D and/or 3-D Convolutional Neural Networks (CNNs) (e.g. VGG, ResNet or C3D) to extract spatial and/or temporal features to encode video contents; and 2) applying Recurrent Neural Networks (RNNs) to generate sentences to describe events in videos. Temporal attention-based model has gained much progress by considering the importance of each video frame. However, for a long video, especially for a video which consists of a set of sub-events, we should discover and leverage the importance of each sub-shot instead of each frame. In this paper, we propose a novel approach, namely temporal and spatial LSTM (TS-LSTM), which systematically exploits spatial and temporal dynamics within video sequences. In TS-LSTM, a temporal pooling LSTM (TP-LSTM) is designed to incorporate both spatial and temporal information to extract long-term temporal dynamics within video sub-shots; and a stacked LSTM is introduced to generate a list of words to describe the video. Experimental results obtained in two public video captioning benchmarks indicate that our TS-LSTM outperforms the state-of-the-art methods.

Yuyu Guo, Lianli Gao, Jingkuan Song et al.

Scene graph generation (SGG) is built on top of detected objects to predict object pairwise visual relations for describing the image content abstraction. Existing works have revealed that if the links between objects are given as prior knowledge, the performance of SGG is significantly improved. Inspired by this observation, in this article, we propose a relation regularized network (R2-Net), which can predict whether there is a relationship between two objects and encode this relation into object feature refinement and better SGG. Specifically, we first construct an affinity matrix among detected objects to represent the probability of a relationship between two objects. Graph convolution networks (GCNs) over this relation affinity matrix are then used as object encoders, producing relation-regularized representations of objects. With these relation-regularized features, our R2-Net can effectively refine object labels and generate scene graphs. Extensive experiments are conducted on the visual genome dataset for three SGG tasks (i.e., predicate classification, scene graph classification, and scene graph detection), demonstrating the effectiveness of our proposed method. Ablation studies also verify the key roles of our proposed components in performance improvement.

Yuyu Guo, Jingkuan Song, Lianli Gao et al.

As a structured representation of the image content, the visual scene graph (visual relationship) acts as a bridge between computer vision and natural language processing. Existing models on the scene graph generation task notoriously require tens or hundreds of labeled samples. By contrast, human beings can learn visual relationships from a few or even one example. Inspired by this, we design a task named One-Shot Scene Graph Generation, where each relationship triplet (e.g., "dog-has-head") comes from only one labeled example. The key insight is that rather than learning from scratch, one can utilize rich prior knowledge. In this paper, we propose Multiple Structured Knowledge (Relational Knowledge and Commonsense Knowledge) for the one-shot scene graph generation task. Specifically, the Relational Knowledge represents the prior knowledge of relationships between entities extracted from the visual content, e.g., the visual relationships "standing in", "sitting in", and "lying in" may exist between "dog" and "yard", while the Commonsense Knowledge encodes "sense-making" knowledge like "dog can guard yard". By organizing these two kinds of knowledge in a graph structure, Graph Convolution Networks (GCNs) are used to extract knowledge-embedded semantic features of the entities. Besides, instead of extracting isolated visual features from each entity generated by Faster R-CNN, we utilize an Instance Relation Transformer encoder to fully explore their context information. Based on a constructed one-shot dataset, the experimental results show that our method significantly outperforms existing state-of-the-art methods by a large margin. Ablation studies also verify the effectiveness of the Instance Relation Transformer encoder and the Multiple Structured Knowledge.

Yuyu Guo, Lei Bi, Dongming Wei et al.

Motion estimation is a fundamental step in dynamic medical image processing for the assessment of target organ anatomy and function. However, existing image-based motion estimation methods, which optimize the motion field by evaluating the local image similarity, are prone to produce implausible estimation, especially in the presence of large motion. In this study, we provide a novel motion estimation framework of Dense-Sparse-Dense (DSD), which comprises two stages. In the first stage, we process the raw dense image to extract sparse landmarks to represent the target organ anatomical topology and discard the redundant information that is unnecessary for motion estimation. For this purpose, we introduce an unsupervised 3D landmark detection network to extract spatially sparse but representative landmarks for the target organ motion estimation. In the second stage, we derive the sparse motion displacement from the extracted sparse landmarks of two images of different time points. Then, we present a motion reconstruction network to construct the motion field by projecting the sparse landmarks displacement back into the dense image domain. Furthermore, we employ the estimated motion field from our two-stage DSD framework as initialization and boost the motion estimation quality in light-weight yet effective iterative optimization. We evaluate our method on two dynamic medical imaging tasks to model cardiac motion and lung respiratory motion, respectively. Our method has produced superior motion estimation accuracy compared to existing comparative methods. Besides, the extensive experimental results demonstrate that our solution can extract well representative anatomical landmarks without any requirement of manual annotation. Our code is publicly available online.

Yuyu Guo, Lianli Gao, Xuanhan Wang et al.

The scene graph generation (SGG) task aims to detect visual relationship triplets, i.e., subject, predicate, object, in an image, providing a structural vision layout for scene understanding. However, current models are stuck in common predicates, e.g., "on" and "at", rather than informative ones, e.g., "standing on" and "looking at", resulting in the loss of precise information and overall performance. If a model only uses "stone on road" rather than "blocking" to describe an image, it is easy to misunderstand the scene. We argue that this phenomenon is caused by two key imbalances between informative predicates and common ones, i.e., semantic space level imbalance and training sample level imbalance. To tackle this problem, we propose BA-SGG, a simple yet effective SGG framework based on balance adjustment but not the conventional distribution fitting. It integrates two components: Semantic Adjustment (SA) and Balanced Predicate Learning (BPL), respectively for adjusting these imbalances. Benefited from the model-agnostic process, our method is easily applied to the state-of-the-art SGG models and significantly improves the SGG performance. Our method achieves 14.3%, 8.0%, and 6.1% higher Mean Recall (mR) than that of the Transformer model at three scene graph generation sub-tasks on Visual Genome, respectively. Codes are publicly available.

Yuyu Guo, Lei Bi, Euijoon Ahn et al.

Dynamic medical imaging is usually limited in application due to the large radiation doses and longer image scanning and reconstruction times. Existing methods attempt to reduce the dynamic sequence by interpolating the volumes between the acquired image volumes. However, these methods are limited to either 2D images and/or are unable to support large variations in the motion between the image volume sequences. In this paper, we present a spatiotemporal volumetric interpolation network (SVIN) designed for 4D dynamic medical images. SVIN introduces dual networks: first is the spatiotemporal motion network that leverages the 3D convolutional neural network (CNN) for unsupervised parametric volumetric registration to derive spatiotemporal motion field from two-image volumes; the second is the sequential volumetric interpolation network, which uses the derived motion field to interpolate image volumes, together with a new regression-based module to characterize the periodic motion cycles in functional organ structures. We also introduce an adaptive multi-scale architecture to capture the volumetric large anatomy motions. Experimental results demonstrated that our SVIN outperformed state-of-the-art temporal medical interpolation methods and natural video interpolation methods that have been extended to support volumetric images. Our ablation study further exemplified that our motion network was able to better represent the large functional motion compared with the state-of-the-art unsupervised medical registration methods.

Yuyu Guo

Implantation of stents into coronary arteries is a common treatment option for patients with cardiovascular disease. Assessment of safety and efficacy of the stent implantation occurs via manual visual inspection of the neointimal coverage from intravascular optical coherence tomography (IVOCT) images. However, such manual assessment requires the detection of thousands of strut points within the stent. This is a challenging, tedious, and time-consuming task because the strut points usually appear as small, irregular shaped objects with inhomogeneous textures, and are often occluded by shadows, artifacts, and vessel walls. Conventional methods based on textures, edge detection, or simple classifiers for automated detection of strut points in IVOCT images have low recall and precision as they are, unable to adequately represent the visual features of the strut point for detection. In this study, we propose a local-global refinement network to integrate local-patch content with global content for strut points detection from IVOCT images. Our method densely detects the potential strut points in local image patches and then refines them according to global appearance constraints to reduce false positives. Our experimental results on a clinical dataset of 7,000 IVOCT images demonstrated that our method outperformed the state-of-the-art methods with a recall of 0.92 and precision of 0.91 for strut points detection.

Lei Bi, Yuyu Guo, Qian Wang et al.

Automated segmentation of the optic cup and disk on retinal fundus images is fundamental for the automated detection / analysis of glaucoma. Traditional segmentation approaches depend heavily upon hand-crafted features and a priori knowledge of the user. As such, these methods are difficult to be adapt to the clinical environment. Recently, deep learning methods based on fully convolutional networks (FCNs) have been successful in resolving segmentation problems. However, the reliance on large annotated training data is problematic when dealing with medical images. If a sufficient amount of annotated training data to cover all possible variations is not available, FCNs do not provide accurate segmentation. In addition, FCNs have a large receptive field in the convolutional layers, and hence produce coarse outputs of boundaries. Hence, we propose a new fully automated method that we refer to as a dual-stage fully convolutional networks (DSFCN). Our approach leverages deep residual architectures and FCNs and learns and infers the location of the optic cup and disk in a step-wise manner with fine-grained details. During training, our approach learns from the training data and the estimated results derived from the previous iteration. The ability to learn from the previous iteration optimizes the learning of the optic cup and the disk boundaries. During testing (prediction), DSFCN uses test (input) images and the estimated probability map derived from previous iterations to gradually improve the segmentation accuracy. Our method achieved an average Dice co-efficient of 0.8488 and 0.9441 for optic cup and disk segmentation and an area under curve (AUC) of 0.9513 for glaucoma detection.

Jingkuan Song, Yuyu Guo, Lianli Gao et al.

Video captioning in essential is a complex natural process, which is affected by various uncertainties stemming from video content, subjective judgment, etc. In this paper we build on the recent progress in using encoder-decoder framework for video captioning and address what we find to be a critical deficiency of the existing methods, that most of the decoders propagate deterministic hidden states. Such complex uncertainty cannot be modeled efficiently by the deterministic models. In this paper, we propose a generative approach, referred to as multi-modal stochastic RNNs networks (MS-RNN), which models the uncertainty observed in the data using latent stochastic variables. Therefore, MS-RNN can improve the performance of video captioning, and generate multiple sentences to describe a video considering different random factors. Specifically, a multi-modal LSTM (M-LSTM) is first proposed to interact with both visual and textual features to capture a high-level representation. Then, a backward stochastic LSTM (S-LSTM) is proposed to support uncertainty propagation by introducing latent variables. Experimental results on the challenging datasets MSVD and MSR-VTT show that our proposed MS-RNN approach outperforms the state-of-the-art video captioning benchmarks.