Yu-Xiong Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yu-Xiong Wang has a total of 129 co-authors. Here are the top co-authors by number of shared publications:
- Martial Hebert
Carnegie Mellon University
5 shared publications - Pavel Tokmakov
Toyota Research Institute
3 shared publications - Zhipeng Bao
Carnegie Mellon University
3 shared publications - Ziqi Pang
University of Illinois Urbana-Champaign
3 shared publications - Jiahua Dong
University of Illinois Urbana-Champaign University of Illinois Urbana-Champaign University of Illinois Urbana-Champaign
2 shared publications - Liang-Yan Gui
University of Illinois Urbana-Champaign
2 shared publications
Latest Publications
- Open Access
Capturing Visual Environment Structure Correlates with Control Performance
The choice of visual representation is key to scaling generalist robot policies. However, direct evaluation via policy rollouts is expensive, even in simulation. Existing proxy metrics focus on the representation's capacity to capture narrow aspects of the visual world, like object shape, limiting generalization across environments. In this paper, we take an analytical perspective: we probe pretrained visual encoders by measuring how well they support decoding of environment state -- including geometry, object structure, and physical attributes -- from images. Leveraging simulation environments with access to ground-truth state, we show that this probing accuracy strongly correlates with downstream policy performance across diverse environments and learning settings, significantly outperforming prior metrics and enabling efficient representation selection. More broadly, our study provides insight into the representational properties that support generalizable manipulation, suggesting that learning to encode the latent physical state of the environment is a promising objective for control.
Open AccessCapturing Visual Environment Structure Correlates with Control Performance
The choice of visual representation is key to scaling generalist robot policies. However, direct evaluation via policy rollouts is expensive, even in simulation. Existing proxy metrics focus on the representation's capacity to capture narrow aspects of the visual world, like object shape, limiting generalization across environments. In this paper, we take an analytical perspective: we probe pretrained visual encoders by measuring how well they support decoding of environment state -- including geometry, object structure, and physical attributes -- from images. Leveraging simulation environments with access to ground-truth state, we show that this probing accuracy strongly correlates with downstream policy performance across diverse environments and learning settings, significantly outperforming prior metrics and enabling efficient representation selection. More broadly, our study provides insight into the representational properties that support generalizable manipulation, suggesting that learning to encode the latent physical state of the environment is a promising objective for control.
- Open Access
SCas4D: Structural Cascaded Optimization for Boosting Persistent 4D Novel View Synthesis
Persistent dynamic scene modeling for tracking and novel-view synthesis remains challenging due to the difficulty of capturing accurate deformations while maintaining computational efficiency. We propose SCas4D, a cascaded optimization framework that leverages structural patterns in 3D Gaussian Splatting for dynamic scenes. The key idea is that real-world deformations often exhibit hierarchical patterns, where groups of Gaussians share similar transformations. By progressively refining deformations from coarse part-level to fine point-level, SCas4D achieves convergence within 100 iterations per time frame and produces results comparable to existing methods with only one-twentieth of the training iterations. The approach also demonstrates effectiveness in self-supervised articulated object segmentation, novel view synthesis, and dense point tracking tasks.
Open AccessSCas4D: Structural Cascaded Optimization for Boosting Persistent 4D Novel View Synthesis
Persistent dynamic scene modeling for tracking and novel-view synthesis remains challenging due to the difficulty of capturing accurate deformations while maintaining computational efficiency. We propose SCas4D, a cascaded optimization framework that leverages structural patterns in 3D Gaussian Splatting for dynamic scenes. The key idea is that real-world deformations often exhibit hierarchical patterns, where groups of Gaussians share similar transformations. By progressively refining deformations from coarse part-level to fine point-level, SCas4D achieves convergence within 100 iterations per time frame and produces results comparable to existing methods with only one-twentieth of the training iterations. The approach also demonstrates effectiveness in self-supervised articulated object segmentation, novel view synthesis, and dense point tracking tasks.
- Open Access
MoReact: Generating Reactive Motion from Textual Descriptions
Modeling and generating human reactions poses a significant challenge with broad applications for computer vision and human-computer interaction. Existing methods either treat multiple individuals as a single entity, directly generating interactions, or rely solely on one person's motion to generate the other's reaction, failing to integrate the rich semantic information that underpins human interactions. Yet, these methods often fall short in adaptive responsiveness, i.e., the ability to accurately respond to diverse and dynamic interaction scenarios. Recognizing this gap, our work introduces an approach tailored to address the limitations of existing models by focusing on text-driven human reaction generation. Our model specifically generates realistic motion sequences for individuals that responding to the other's actions based on a descriptive text of the interaction scenario. The goal is to produce motion sequences that not only complement the opponent's movements but also semantically fit the described interactions. To achieve this, we present MoReact, a diffusion-based method designed to disentangle the generation of global trajectories and local motions sequentially. This approach stems from the observation that generating global trajectories first is crucial for guiding local motion, ensuring better alignment with given action and text. Furthermore, we introduce a novel interaction loss to enhance the realism of generated close interactions. Our experiments, utilizing data adapted from a two-person motion dataset, demonstrate the efficacy of our approach for this novel task, which is capable of producing realistic, diverse, and controllable reactions that not only closely match the movements of the counterpart but also adhere to the textual guidance. Please find our webpage at https://xiyan-xu.github.io/MoReactWebPage.
Open AccessMoReact: Generating Reactive Motion from Textual Descriptions
Modeling and generating human reactions poses a significant challenge with broad applications for computer vision and human-computer interaction. Existing methods either treat multiple individuals as a single entity, directly generating interactions, or rely solely on one person's motion to generate the other's reaction, failing to integrate the rich semantic information that underpins human interactions. Yet, these methods often fall short in adaptive responsiveness, i.e., the ability to accurately respond to diverse and dynamic interaction scenarios. Recognizing this gap, our work introduces an approach tailored to address the limitations of existing models by focusing on text-driven human reaction generation. Our model specifically generates realistic motion sequences for individuals that responding to the other's actions based on a descriptive text of the interaction scenario. The goal is to produce motion sequences that not only complement the opponent's movements but also semantically fit the described interactions. To achieve this, we present MoReact, a diffusion-based method designed to disentangle the generation of global trajectories and local motions sequentially. This approach stems from the observation that generating global trajectories first is crucial for guiding local motion, ensuring better alignment with given action and text. Furthermore, we introduce a novel interaction loss to enhance the realism of generated close interactions. Our experiments, utilizing data adapted from a two-person motion dataset, demonstrate the efficacy of our approach for this novel task, which is capable of producing realistic, diverse, and controllable reactions that not only closely match the movements of the counterpart but also adhere to the textual guidance. Please find our webpage at https://xiyan-xu.github.io/MoReactWebPage.
- 12 May 2025
- Published by Springer Nature
- p. 357-383
The Second Visual Object Tracking Segmentation VOTS2024 Challenge Results
- 12 May 2025
- Published by Springer Nature
- p. 357-383
The Second Visual Object Tracking Segmentation VOTS2024 Challenge Results
- 17 November 2024
- Published by Springer Nature
- p. 384-400
AlignDiff: Aligning Diffusion Models for General Few-Shot Segmentation
- 17 November 2024
- Published by Springer Nature
- p. 384-400
AlignDiff: Aligning Diffusion Models for General Few-Shot Segmentation
- Open Access
- 13 November 2024
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 133 (5) , 2247-2258
Learning to Detect Novel Species with SAM in the Wild
This paper tackles the limitation of a closed-world object detection model that was trained on one species. The expectation for this model is that it will not generalize well to recognize the instances of new species if they were present in the incoming data stream. We propose a novel object detection framework for this open-world setting that is suitable for applications that monitor wildlife, ocean life, livestock, plant phenotype and crops that typically feature one species in the image. Our method leverages labeled samples from one species in combination with a novelty detection method and Segment Anything Model, a vision foundation model, to (1) identify the presence of new species in unlabeled images, (2) localize their instances, and (3) retrain the initial model with the localized novel class instances. The resulting integrated system assimilates and learns from unlabeled samples of the new classes while not “forgetting” the original species the model was trained on. We demonstrate our findings on two different domains, (1) wildlife detection and (2) plant detection. Our method achieves an AP of 56.2 (for 4 novel species) to 61.6 (for 1 novel species) for wildlife domain, without relying on any ground truth data in the background.
- 13 November 2024
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 133 (5) , 2247-2258
Open AccessLearning to Detect Novel Species with SAM in the Wild
This paper tackles the limitation of a closed-world object detection model that was trained on one species. The expectation for this model is that it will not generalize well to recognize the instances of new species if they were present in the incoming data stream. We propose a novel object detection framework for this open-world setting that is suitable for applications that monitor wildlife, ocean life, livestock, plant phenotype and crops that typically feature one species in the image. Our method leverages labeled samples from one species in combination with a novelty detection method and Segment Anything Model, a vision foundation model, to (1) identify the presence of new species in unlabeled images, (2) localize their instances, and (3) retrain the initial model with the localized novel class instances. The resulting integrated system assimilates and learns from unlabeled samples of the new classes while not “forgetting” the original species the model was trained on. We demonstrate our findings on two different domains, (1) wildlife detection and (2) plant detection. Our method achieves an AP of 56.2 (for 4 novel species) to 61.6 (for 1 novel species) for wildlife domain, without relying on any ground truth data in the background.
- Open Access
Separate-and-Enhance: Compositional Finetuning for Text2Image Diffusion Models
Despite recent significant strides achieved by diffusion-based Text-to-Image (T2I) models, current systems are still less capable of ensuring decent compositional generation aligned with text prompts, particularly for the multi-object generation. This work illuminates the fundamental reasons for such misalignment, pinpointing issues related to low attention activation scores and mask overlaps. While previous research efforts have individually tackled these issues, we assert that a holistic approach is paramount. Thus, we propose two novel objectives, the Separate loss and the Enhance loss, that reduce object mask overlaps and maximize attention scores, respectively. Our method diverges from conventional test-time-adaptation techniques, focusing on finetuning critical parameters, which enhances scalability and generalizability. Comprehensive evaluations demonstrate the superior performance of our model in terms of image realism, text-image alignment, and adaptability, notably outperforming prominent baselines. Ultimately, this research paves the way for T2I diffusion models with enhanced compositional capacities and broader applicability.
Open AccessSeparate-and-Enhance: Compositional Finetuning for Text2Image Diffusion Models
Despite recent significant strides achieved by diffusion-based Text-to-Image (T2I) models, current systems are still less capable of ensuring decent compositional generation aligned with text prompts, particularly for the multi-object generation. This work illuminates the fundamental reasons for such misalignment, pinpointing issues related to low attention activation scores and mask overlaps. While previous research efforts have individually tackled these issues, we assert that a holistic approach is paramount. Thus, we propose two novel objectives, the Separate loss and the Enhance loss, that reduce object mask overlaps and maximize attention scores, respectively. Our method diverges from conventional test-time-adaptation techniques, focusing on finetuning critical parameters, which enhances scalability and generalizability. Comprehensive evaluations demonstrate the superior performance of our model in terms of image realism, text-image alignment, and adaptability, notably outperforming prominent baselines. Ultimately, this research paves the way for T2I diffusion models with enhanced compositional capacities and broader applicability.
- Open Access
MV-Map: Offboard HD-Map Generation with Multi-view Consistency
While bird's-eye-view (BEV) perception models can be useful for building high-definition maps (HD-Maps) with less human labor, their results are often unreliable and demonstrate noticeable inconsistencies in the predicted HD-Maps from different viewpoints. This is because BEV perception is typically set up in an 'onboard' manner, which restricts the computation and consequently prevents algorithms from reasoning multiple views simultaneously. This paper overcomes these limitations and advocates a more practical 'offboard' HD-Map generation setup that removes the computation constraints, based on the fact that HD-Maps are commonly reusable infrastructures built offline in data centers. To this end, we propose a novel offboard pipeline called MV-Map that capitalizes multi-view consistency and can handle an arbitrary number of frames with the key design of a 'region-centric' framework. In MV-Map, the target HD-Maps are created by aggregating all the frames of onboard predictions, weighted by the confidence scores assigned by an 'uncertainty network'. To further enhance multi-view consistency, we augment the uncertainty network with the global 3D structure optimized by a voxelized neural radiance field (Voxel-NeRF). Extensive experiments on nuScenes show that our MV-Map significantly improves the quality of HD-Maps, further highlighting the importance of offboard methods for HD-Map generation.
Open AccessMV-Map: Offboard HD-Map Generation with Multi-view Consistency
While bird's-eye-view (BEV) perception models can be useful for building high-definition maps (HD-Maps) with less human labor, their results are often unreliable and demonstrate noticeable inconsistencies in the predicted HD-Maps from different viewpoints. This is because BEV perception is typically set up in an 'onboard' manner, which restricts the computation and consequently prevents algorithms from reasoning multiple views simultaneously. This paper overcomes these limitations and advocates a more practical 'offboard' HD-Map generation setup that removes the computation constraints, based on the fact that HD-Maps are commonly reusable infrastructures built offline in data centers. To this end, we propose a novel offboard pipeline called MV-Map that capitalizes multi-view consistency and can handle an arbitrary number of frames with the key design of a 'region-centric' framework. In MV-Map, the target HD-Maps are created by aggregating all the frames of onboard predictions, weighted by the confidence scores assigned by an 'uncertainty network'. To further enhance multi-view consistency, we augment the uncertainty network with the global 3D structure optimized by a voxelized neural radiance field (Voxel-NeRF). Extensive experiments on nuScenes show that our MV-Map significantly improves the quality of HD-Maps, further highlighting the importance of offboard methods for HD-Map generation.
- 1 October 2023
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 66 (2) , 1205-1229
Generalized few-shot node classification: toward an uncertainty-based solution
- 1 October 2023
- Published by Springer Nature in Knowledge and Information Systems
- Vol. 66 (2) , 1205-1229
Generalized few-shot node classification: toward an uncertainty-based solution
- Open Access
Revisiting Deformable Convolution for Depth Completion
Depth completion, which aims to generate high-quality dense depth maps from sparse depth maps, has attracted increasing attention in recent years. Previous work usually employs RGB images as guidance, and introduces iterative spatial propagation to refine estimated coarse depth maps. However, most of the propagation refinement methods require several iterations and suffer from a fixed receptive field, which may contain irrelevant and useless information with very sparse input. In this paper, we address these two challenges simultaneously by revisiting the idea of deformable convolution. We propose an effective architecture that leverages deformable kernel convolution as a single-pass refinement module, and empirically demonstrate its superiority. To better understand the function of deformable convolution and exploit it for depth completion, we further systematically investigate a variety of representative strategies. Our study reveals that, different from prior work, deformable convolution needs to be applied on an estimated depth map with a relatively high density for better performance. We evaluate our model on the large-scale KITTI dataset and achieve state-of-the-art level performance in both accuracy and inference speed. Our code is available at https://github.com/AlexSunNik/ReDC.
Open AccessRevisiting Deformable Convolution for Depth Completion
Depth completion, which aims to generate high-quality dense depth maps from sparse depth maps, has attracted increasing attention in recent years. Previous work usually employs RGB images as guidance, and introduces iterative spatial propagation to refine estimated coarse depth maps. However, most of the propagation refinement methods require several iterations and suffer from a fixed receptive field, which may contain irrelevant and useless information with very sparse input. In this paper, we address these two challenges simultaneously by revisiting the idea of deformable convolution. We propose an effective architecture that leverages deformable kernel convolution as a single-pass refinement module, and empirically demonstrate its superiority. To better understand the function of deformable convolution and exploit it for depth completion, we further systematically investigate a variety of representative strategies. Our study reveals that, different from prior work, deformable convolution needs to be applied on an estimated depth map with a relatively high density for better performance. We evaluate our model on the large-scale KITTI dataset and achieve state-of-the-art level performance in both accuracy and inference speed. Our code is available at https://github.com/AlexSunNik/ReDC.