Fan Lu

Zehan Zheng, Danni Wu, Ruisi Lu et al.

In recent years, there has been a significant increase in focus on the interpolation task of computer vision. Despite the tremendous advancement of video interpolation, point cloud interpolation remains insufficiently explored. Meanwhile, the existence of numerous nonlinear large motions in real-world scenarios makes the point cloud interpolation task more challenging. In light of these issues, we present NeuralPCI: an end-to-end 4D spatio-temporal Neural field for 3D Point Cloud Interpolation, which implicitly integrates multi-frame information to handle nonlinear large motions for both indoor and outdoor scenarios. Furthermore, we construct a new multi-frame point cloud interpolation dataset called NL-Drive for large nonlinear motions in autonomous driving scenes to better demonstrate the superiority of our method. Ultimately, NeuralPCI achieves state-of-the-art performance on both DHB (Dynamic Human Bodies) and NL-Drive datasets. Beyond the interpolation task, our method can be naturally extended to point cloud extrapolation, morphing, and auto-labeling, which indicates its substantial potential in other domains. Codes are available at https://github.com/ispc-lab/NeuralPCI.

Fan Lu, Kai Zhu, Wei Zhai et al.

Semantically coherent out-of-distribution (SCOOD) detection aims to discern outliers from the intended data distribution with access to unlabeled extra set. The coexistence of in-distribution and out-of-distribution samples will exacerbate the model overfitting when no distinction is made. To address this problem, we propose a novel uncertainty-aware optimal transport scheme. Our scheme consists of an energy-based transport (ET) mechanism that estimates the fluctuating cost of uncertainty to promote the assignment of semantic-agnostic representation, and an inter-cluster extension strategy that enhances the discrimination of semantic property among different clusters by widening the corresponding margin distance. Furthermore, a T-energy score is presented to mitigate the magnitude gap between the parallel transport and classifier branches. Extensive experiments on two standard SCOOD benchmarks demonstrate the above-par OOD detection performance, outperforming the state-of-the-art methods by a margin of 27.69% and 34.4% on FPR@95, respectively.

Fan Lu, Sean Meyn

The paper introduces the first formulation of convex Q-learning for Markov decision processes with function approximation. The algorithms and theory rest on a relaxation of a dual of Manne's celebrated linear programming characterization of optimal control. The main contributions firstly concern properties of the relaxation, described as a deterministic convex program: we identify conditions for a bounded solution, and a significant relationship between the solution to the new convex program, and the solution to standard Q-learning. The second set of contributions concern algorithm design and analysis: (i) A direct model-free method for approximating the convex program for Q-learning shares properties with its ideal. In particular, a bounded solution is ensured subject to a simple property of the basis functions; (ii) The proposed algorithms are convergent and new techniques are introduced to obtain the rate of convergence in a mean-square sense; (iii) The approach can be generalized to a range of performance criteria, and it is found that variance can be reduced by considering ``relative'' dynamic programming equations; (iv) The theory is illustrated with an application to a classical inventory control problem.

Fan Lu, Yan Xu, Guang Chen et al.

Neural Radiance Fields (NeRFs) have achieved great success in the past few years. However, most current methods still require intensive resources due to ray marching-based rendering. To construct urban-level radiance fields efficiently, we design Deformable Neural Mesh Primitive~(DNMP), and propose to parameterize the entire scene with such primitives. The DNMP is a flexible and compact neural variant of classic mesh representation, which enjoys both the efficiency of rasterization-based rendering and the powerful neural representation capability for photo-realistic image synthesis. Specifically, a DNMP consists of a set of connected deformable mesh vertices with paired vertex features to parameterize the geometry and radiance information of a local area. To constrain the degree of freedom for optimization and lower the storage budgets, we enforce the shape of each primitive to be decoded from a relatively low-dimensional latent space. The rendering colors are decoded from the vertex features (interpolated with rasterization) by a view-dependent MLP. The DNMP provides a new paradigm for urban-level scene representation with appealing properties: $(1)$ High-quality rendering. Our method achieves leading performance for novel view synthesis in urban scenarios. $(2)$ Low computational costs. Our representation enables fast rendering (2.07ms/1k pixels) and low peak memory usage (110MB/1k pixels). We also present a lightweight version that can run 33$\times$ faster than vanilla NeRFs, and comparable to the highly-optimized Instant-NGP (0.61 vs 0.71ms/1k pixels). Project page: \href{https://dnmp.github.io/}{https://dnmp.github.io/}.

Fan Lu, Joel Mathias, Sean Meyn et al.

Convex Q-learning is a recent approach to reinforcement learning, motivated by the possibility of a firmer theory for convergence, and the possibility of making use of greater a priori knowledge regarding policy or value function structure. This paper explores algorithm design in the continuous time domain, with finite-horizon optimal control objective. The main contributions are (i) Algorithm design is based on a new Q-ODE, which defines the model-free characterization of the Hamilton-Jacobi-Bellman equation. (ii) The Q-ODE motivates a new formulation of Convex Q-learning that avoids the approximations appearing in prior work. The Bellman error used in the algorithm is defined by filtered measurements, which is beneficial in the presence of measurement noise. (iii) A characterization of boundedness of the constraint region is obtained through a non-trivial extension of recent results from the discrete time setting. (iv) The theory is illustrated in application to resource allocation for distributed energy resources, for which the theory is ideally suited.

Fan Lu, Prashant Mehta, Sean Meyn et al.

In recent years there has been a collective research effort to find new formulations of reinforcement learning that are simultaneously more efficient and more amenable to analysis. This paper concerns one approach that builds on the linear programming (LP) formulation of optimal control of Manne. A primal version is called logistic Q-learning, and a dual variant is convex Q-learning. This paper focuses on the latter, while building bridges with the former. The main contributions follow: (i) The dual of convex Q-learning is not precisely Manne's LP or a version of logistic Q-learning, but has similar structure that reveals the need for regularization to avoid over-fitting. (ii) A sufficient condition is obtained for a bounded solution to the Q-learning LP. (iii) Simulation studies reveal numerical challenges when addressing sampled-data systems based on a continuous time model. The challenge is addressed using state-dependent sampling. The theory is illustrated with applications to examples from OpenAI gym. It is shown that convex Q-learning is successful in cases where standard Q-learning diverges, such as the LQR problem.

Chenfeng Wang, Wei He, Xuhan Zhu et al.

In language reasoning, longer chains of thought consistently yield better performance, which naturally suggests that visual latent reasoning may likewise benefit from longer latent sequences. However, we discover a counterintuitive phenomenon: the performance of existing latent visual reasoning methods systematically degrades as the latent sequence grows longer. We reveal the root cause: Information Gain Collapse -- autoregressive generation makes each step highly dependent on prior outputs, so subsequent tokens can barely introduce new information. We further identify that heavily pooled ($\geq 128\times$) image embeddings used as supervision targets provide no more signal than meaningless placeholders. Motivated by these insights, we propose SCOLAR (Self-COnsistent LAtent Reasoning), which introduces a lightweight detransformer that leverages the LLM's full-sequence hidden states to generate auxiliary visual tokens in a single shot, with each token independently anchored to the original visual space. Combined with three-stage SFT and ALPO reinforcement learning, SCOLAR extends acceptable latent CoT length by over $30\times$, achieves state-of-the-art among open-source models on real-world reasoning benchmarks (+14.12% over backbone), and demonstrates strong out-of-distribution generalization.

Weiyi Xue, Fan Lu, Guang Chen

Outdoor LiDAR point clouds are typically large-scale and complexly distributed. To achieve efficient and accurate registration, emphasizing the similarity among local regions and prioritizing global local-to-local matching is of utmost importance, subsequent to which accuracy can be enhanced through cost-effective fine registration. In this paper, a novel hierarchical neural network with double attention named HDMNet is proposed for large-scale outdoor LiDAR point cloud registration. Specifically, A novel feature consistency enhanced double-soft matching network is introduced to achieve two-stage matching with high flexibility while enlarging the receptive field with high efficiency in a patch-to patch manner, which significantly improves the registration performance. Moreover, in order to further utilize the sparse matching information from deeper layer, we develop a novel trainable embedding mask to incorporate the confidence scores of correspondences obtained from pose estimation of deeper layer, eliminating additional computations. The high-confidence keypoints in the sparser point cloud of the deeper layer correspond to a high-confidence spatial neighborhood region in shallower layer, which will receive more attention, while the features of non-key regions will be masked. Extensive experiments are conducted on two large-scale outdoor LiDAR point cloud datasets to demonstrate the high accuracy and efficiency of the proposed HDMNet.

Weiyi Xue, Zehan Zheng, Fan Lu et al.

Although recent efforts have extended Neural Radiance Fields (NeRF) into LiDAR point cloud synthesis, the majority of existing works exhibit a strong dependence on precomputed poses. However, point cloud registration methods struggle to achieve precise global pose estimation, whereas previous pose-free NeRFs overlook geometric consistency in global reconstruction. In light of this, we explore the geometric insights of point clouds, which provide explicit registration priors for reconstruction. Based on this, we propose Geometry guided Neural LiDAR Fields(GeoNLF), a hybrid framework performing alternately global neural reconstruction and pure geometric pose optimization. Furthermore, NeRFs tend to overfit individual frames and easily get stuck in local minima under sparse-view inputs. To tackle this issue, we develop a selective-reweighting strategy and introduce geometric constraints for robust optimization. Extensive experiments on NuScenes and KITTI-360 datasets demonstrate the superiority of GeoNLF in both novel view synthesis and multi-view registration of low-frequency large-scale point clouds.

Wenjie Mu, Zhan Li, Chuanzhou Su et al.

Generalizable Neural Radiance Fields (GeNeRFs) enable high-quality scene reconstruction from sparse views and can generalize to unseen scenes. However, in real-world settings, transient distractors break cross-view structural consistency, corrupting supervision and degrading reconstruction quality. Existing distractor-free NeRF methods rely on per-scene optimization and estimate uncertainty from per-view reconstruction errors, which are not reliable for GeNeRFs and often misjudge inconsistent static structures as distractors. To this end, we propose MU-GeNeRF, a Multi-view Uncertainty-guided distractor-aware GeNeRF framework designed to alleviate GeNeRF's robust modeling challenges in the presence of transient distractions. We decompose distractor awareness into two complementary uncertainty components: Source-view Uncertainty, which captures structural discrepancies across source views caused by viewpoint changes or dynamic factors; and Target-view Uncertainty, which detects observation anomalies in the target image induced by transient distractors.These two uncertainties address distinct error sources and are combined through a heteroscedastic reconstruction loss, which guides the model to adaptively modulate supervision, enabling more robust distractor suppression and geometric modeling.Extensive experiments show that our method not only surpasses existing GeNeRFs but also achieves performance comparable to scene-specific distractor-free NeRFs.

Zehan Zheng, Fan Lu, Weiyi Xue et al.

Although neural radiance fields (NeRFs) have achieved triumphs in image novel view synthesis (NVS), LiDAR NVS remains largely unexplored. Previous LiDAR NVS methods employ a simple shift from image NVS methods while ignoring the dynamic nature and the large-scale reconstruction problem of LiDAR point clouds. In light of this, we propose LiDAR4D, a differentiable LiDAR-only framework for novel space-time LiDAR view synthesis. In consideration of the sparsity and large-scale characteristics, we design a 4D hybrid representation combined with multi-planar and grid features to achieve effective reconstruction in a coarse-to-fine manner. Furthermore, we introduce geometric constraints derived from point clouds to improve temporal consistency. For the realistic synthesis of LiDAR point clouds, we incorporate the global optimization of ray-drop probability to preserve cross-region patterns. Extensive experiments on KITTI-360 and NuScenes datasets demonstrate the superiority of our method in accomplishing geometry-aware and time-consistent dynamic reconstruction. Codes are available at https://github.com/ispc-lab/LiDAR4D.

Wei Wu, Fan Lu, Yunnan Wang et al.

Offering great potential in robotic manipulation, a capable Vision-Language-Action (VLA) foundation model is expected to faithfully generalize across tasks and platforms while ensuring cost efficiency (e.g., data and GPU hours required for adaptation). To this end, we develop LingBot-VLA with around 20,000 hours of real-world data from 9 popular dual-arm robot configurations. Through a systematic assessment on 3 robotic platforms, each completing 100 tasks with 130 post-training episodes per task, our model achieves clear superiority over competitors, showcasing its strong performance and broad generalizability. We have also built an efficient codebase, which delivers a throughput of 261 samples per second per GPU with an 8-GPU training setup, representing a 1.5~2.8$\times$ (depending on the relied VLM base model) speedup over existing VLA-oriented codebases. The above features ensure that our model is well-suited for real-world deployment. To advance the field of robot learning, we provide open access to the code, base model, and benchmark data, with a focus on enabling more challenging tasks and promoting sound evaluation standards.

Hang Ye, Xiaoxuan Ma, Fan Lu et al.

Digital human generation has been studied for decades and supports a wide range of real-world applications. However, most existing systems are passively animated, relying on privileged state or scripted control, which limits scalability to novel environments. We instead ask: how can digital humans actively behave using only visual observations and specified goals in novel scenes? Achieving this would enable populating any 3D environments with digital humans at scale that exhibit spontaneous, natural, goal-directed behaviors. To this end, we introduce Visually-grounded Humanoid Agents, a coupled two-layer (world-agent) paradigm that replicates humans at multiple levels: they look, perceive, reason, and behave like real people in real-world 3D scenes. The World Layer reconstructs semantically rich 3D Gaussian scenes from real-world videos via an occlusion-aware pipeline and accommodates animatable Gaussian-based human avatars. The Agent Layer transforms these avatars into autonomous humanoid agents, equipping them with first-person RGB-D perception and enabling them to perform accurate, embodied planning with spatial awareness and iterative reasoning, which is then executed at the low level as full-body actions to drive their behaviors in the scene. We further introduce a benchmark to evaluate humanoid-scene interaction in diverse reconstructed environments. Experiments show our agents achieve robust autonomous behavior, yielding higher task success rates and fewer collisions than ablations and state-of-the-art planning methods. This work enables active digital human population and advances human-centric embodied AI. Data, code, and models will be open-sourced.

Haiyun Wei, Fan Lu, Yunwei Zhu et al.

Generating realistic and diverse LiDAR point clouds is crucial for autonomous driving simulation. Although previous methods achieve LiDAR point cloud generation from user inputs, they struggle to attain high-quality results while enabling versatile controllability, due to the imbalance between the complex distribution of LiDAR point clouds and the simple control signals. To address the limitation, we propose LiDARDraft, which utilizes the 3D layout to build a bridge between versatile conditional signals and LiDAR point clouds. The 3D layout can be trivially generated from various user inputs such as textual descriptions and images. Specifically, we represent text, images, and point clouds as unified 3D layouts, which are further transformed into semantic and depth control signals. Then, we employ a rangemap-based ControlNet to guide LiDAR point cloud generation. This pixel-level alignment approach demonstrates excellent performance in controllable LiDAR point clouds generation, enabling "simulation from scratch", allowing self-driving environments to be created from arbitrary textual descriptions, images and sketches.

Fan Lu, Qimai Li, Bo Liu et al.

User preference modeling is a vital yet challenging problem in personalized product search. In recent years, latent space based methods have achieved state-of-the-art performance by jointly learning semantic representations of products, users, and text tokens. However, existing methods are limited in their ability to model user preferences. They typically represent users by the products they visited in a short span of time using attentive models and lack the ability to exploit relational information such as user-product interactions or item co-occurrence relations. In this work, we propose to address the limitations of prior arts by exploring local and global user behavior patterns on a user successive behavior graph, which is constructed by utilizing short-term actions of all users. To capture implicit user preference signals and collaborative patterns, we use an efficient jumping graph convolution to explore high-order relations to enrich product representations for user preference modeling. Our approach can be seamlessly integrated with existing latent space based methods and be potentially applied in any product retrieval method that uses purchase history to model user preferences. Extensive experiments on eight Amazon benchmarks demonstrate the effectiveness and potential of our approach. The source code is available at \url{https://github.com/floatSDSDS/SBG}.

Sanqing Qu, Guang Chen, Zhijun Li et al.

Weakly-supervised temporal action localization aims to localize action instances temporal boundary and identify the corresponding action category with only video-level labels. Traditional methods mainly focus on foreground and background frames separation with only a single attention branch and class activation sequence. However, we argue that apart from the distinctive foreground and background frames there are plenty of semantically ambiguous action context frames. It does not make sense to group those context frames to the same background class since they are semantically related to a specific action category. Consequently, it is challenging to suppress action context frames with only a single class activation sequence. To address this issue, in this paper, we propose an action-context modeling network termed ACM-Net, which integrates a three-branch attention module to measure the likelihood of each temporal point being action instance, context, or non-action background, simultaneously. Then based on the obtained three-branch attention values, we construct three-branch class activation sequences to represent the action instances, contexts, and non-action backgrounds, individually. To evaluate the effectiveness of our ACM-Net, we conduct extensive experiments on two benchmark datasets, THUMOS-14 and ActivityNet-1.3. The experiments show that our method can outperform current state-of-the-art methods, and even achieve comparable performance with fully-supervised methods. Code can be found at https://github.com/ispc-lab/ACM-Net

Fan Lu, Guang Chen, Sanqing Qu et al.

LiDAR point cloud streams are usually sparse in time dimension, which is limited by hardware performance. Generally, the frame rates of mechanical LiDAR sensors are 10 to 20 Hz, which is much lower than other commonly used sensors like cameras. To overcome the temporal limitations of LiDAR sensors, a novel task named Point Cloud Frame Interpolation is studied in this paper. Given two consecutive point cloud frames, Point Cloud Frame Interpolation aims to generate intermediate frame(s) between them. To achieve that, we propose a novel framework, namely Point Cloud Frame Interpolation Network (PointINet). Based on the proposed method, the low frame rate point cloud streams can be upsampled to higher frame rates. We start by estimating bi-directional 3D scene flow between the two point clouds and then warp them to the given time step based on the 3D scene flow. To fuse the two warped frames and generate intermediate point cloud(s), we propose a novel learning-based points fusion module, which simultaneously takes two warped point clouds into consideration. We design both quantitative and qualitative experiments to evaluate the performance of the point cloud frame interpolation method and extensive experiments on two large scale outdoor LiDAR datasets demonstrate the effectiveness of the proposed PointINet. Our code is available at https://github.com/ispc-lab/PointINet.git.

Fan Lu, Guang Chen, Yinlong Liu et al.

Keypoint detector and descriptor are two main components of point cloud registration. Previous learning-based keypoint detectors rely on saliency estimation for each point or farthest point sample (FPS) for candidate points selection, which are inefficient and not applicable in large scale scenes. This paper proposes Random Sample-based Keypoint Detector and Descriptor Network (RSKDD-Net) for large scale point cloud registration. The key idea is using random sampling to efficiently select candidate points and using a learning-based method to jointly generate keypoints and descriptors. To tackle the information loss of random sampling, we exploit a novel random dilation cluster strategy to enlarge the receptive field of each sampled point and an attention mechanism to aggregate the positions and features of neighbor points. Furthermore, we propose a matching loss to train the descriptor in a weakly supervised manner. Extensive experiments on two large scale outdoor LiDAR datasets show that the proposed RSKDD-Net achieves state-of-the-art performance with more than 15 times faster than existing methods. Our code is available at https://github.com/ispc-lab/RSKDD-Net.

Kecheng Zheng, Yifei Zhang, Wei Wu et al.

Language-image pre-training largely relies on how precisely and thoroughly a text describes its paired image. In practice, however, the contents of an image can be so rich that well describing them requires lengthy captions (e.g., with 10 sentences), which are usually missing in existing datasets. Consequently, there are currently no clear evidences on whether and how language-image pre-training could benefit from long captions. To figure this out, we first re-caption 30M images with detailed descriptions using a pre-trained Multi-modality Large Language Model (MLLM), and then study the usage of the resulting captions under a contrastive learning framework. We observe that, each sentence within a long caption is very likely to describe the image partially (e.g., an object). Motivated by this, we propose to dynamically sample sub-captions from the text label to construct multiple positive pairs, and introduce a grouping loss to match the embeddings of each sub-caption with its corresponding local image patches in a self-supervised manner. Experimental results on a wide rage of downstream tasks demonstrate the consistent superiority of our method, termed DreamLIP, over previous alternatives, highlighting its fine-grained representational capacity. It is noteworthy that, on the tasks of image-text retrieval and semantic segmentation, our model trained with 30M image-text pairs achieves on par or even better performance than CLIP trained with 400M pairs. Project page is available at https://zyf0619sjtu.github.io/dream-lip.

Wenyao Zhang, Hongsi Liu, Zekun Qi et al.

Recent advances in vision-language-action (VLA) models have shown promise in integrating image generation with action prediction to improve generalization and reasoning in robot manipulation. However, existing methods are limited to challenging image-based forecasting, which suffers from redundant information and lacks comprehensive and critical world knowledge, including dynamic, spatial and semantic information. To address these limitations, we propose DreamVLA, a novel VLA framework that integrates comprehensive world knowledge forecasting to enable inverse dynamics modeling, thereby establishing a perception-prediction-action loop for manipulation tasks. Specifically, DreamVLA introduces a dynamic-region-guided world knowledge prediction, integrated with the spatial and semantic cues, which provide compact yet comprehensive representations for action planning. This design aligns with how humans interact with the world by first forming abstract multimodal reasoning chains before acting. To mitigate interference among the dynamic, spatial and semantic information during training, we adopt a block-wise structured attention mechanism that masks their mutual attention, preventing information leakage and keeping each representation clean and disentangled. Moreover, to model the conditional distribution over future actions, we employ a diffusion-based transformer that disentangles action representations from shared latent features. Extensive experiments on both real-world and simulation environments demonstrate that DreamVLA achieves 76.7% success rate on real robot tasks and 4.44 average length on the CALVIN ABC-D benchmarks.

Fan Lu, Wei Wu, Kecheng Zheng et al.

Generating detailed captions comprehending text-rich visual content in images has received growing attention for Large Vision-Language Models (LVLMs). However, few studies have developed benchmarks specifically tailored for detailed captions to measure their accuracy and comprehensiveness. In this paper, we introduce a detailed caption benchmark, termed as CompreCap, to evaluate the visual context from a directed scene graph view. Concretely, we first manually segment the image into semantically meaningful regions (i.e., semantic segmentation mask) according to common-object vocabulary, while also distinguishing attributes of objects within all those regions. Then directional relation labels of these objects are annotated to compose a directed scene graph that can well encode rich compositional information of the image. Based on our directed scene graph, we develop a pipeline to assess the generated detailed captions from LVLMs on multiple levels, including the object-level coverage, the accuracy of attribute descriptions, the score of key relationships, etc. Experimental results on the CompreCap dataset confirm that our evaluation method aligns closely with human evaluation scores across LVLMs.

Fan Lu, Kwan-Yee Lin, Yan Xu et al.

Text-to-3D generation has achieved remarkable success via large-scale text-to-image diffusion models. Nevertheless, there is no paradigm for scaling up the methodology to urban scale. Urban scenes, characterized by numerous elements, intricate arrangement relationships, and vast scale, present a formidable barrier to the interpretability of ambiguous textual descriptions for effective model optimization. In this work, we surmount the limitations by introducing a compositional 3D layout representation into text-to-3D paradigm, serving as an additional prior. It comprises a set of semantic primitives with simple geometric structures and explicit arrangement relationships, complementing textual descriptions and enabling steerable generation. Upon this, we propose two modifications -- (1) We introduce Layout-Guided Variational Score Distillation to address model optimization inadequacies. It conditions the score distillation sampling process with geometric and semantic constraints of 3D layouts. (2) To handle the unbounded nature of urban scenes, we represent 3D scene with a Scalable Hash Grid structure, incrementally adapting to the growing scale of urban scenes. Extensive experiments substantiate the capability of our framework to scale text-to-3D generation to large-scale urban scenes that cover over 1000m driving distance for the first time. We also present various scene editing demonstrations, showing the powers of steerable urban scene generation. Website: https://urbanarchitect.github.io.

Fan Lu, Kai Zhu, Kecheng Zheng et al.

Full-spectrum out-of-distribution (F-OOD) detection aims to accurately recognize in-distribution (ID) samples while encountering semantic and covariate shifts simultaneously. However, existing out-of-distribution (OOD) detectors tend to overfit the covariance information and ignore intrinsic semantic correlation, inadequate for adapting to complex domain transformations. To address this issue, we propose a Likelihood-Aware Semantic Alignment (LSA) framework to promote the image-text correspondence into semantically high-likelihood regions. LSA consists of an offline Gaussian sampling strategy which efficiently samples semantic-relevant visual embeddings from the class-conditional Gaussian distribution, and a bidirectional prompt customization mechanism that adjusts both ID-related and negative context for discriminative ID/OOD boundary. Extensive experiments demonstrate the remarkable OOD detection performance of our proposed LSA especially on the intractable Near-OOD setting, surpassing existing methods by a margin of $15.26\%$ and $18.88\%$ on two F-OOD benchmarks, respectively.

Shuailei Ma, Kecheng Zheng, Ying Wei et al.

Although text-to-image (T2I) models have recently thrived as visual generative priors, their reliance on high-quality text-image pairs makes scaling up expensive. We argue that grasping the cross-modality alignment is not a necessity for a sound visual generative prior, whose focus should be on texture modeling. Such a philosophy inspires us to study image-to-image (I2I) generation, where models can learn from in-the-wild images in a self-supervised manner. We first develop a pure vision-based training framework, Lumos, and confirm the feasibility and the scalability of learning I2I models. We then find that, as an upstream task of T2I, our I2I model serves as a more foundational visual prior and achieves on-par or better performance than existing T2I models using only 1/10 text-image pairs for fine-tuning. We further demonstrate the superiority of I2I priors over T2I priors on some text-irrelevant visual generative tasks, like image-to-3D and image-to-video. Our project page is available at https://ant-research.github.io/lumos.

Haotian Liu, Sanqing Qu, Fan Lu et al.

Event cameras can record scene dynamics with high temporal resolution, providing rich scene details for monocular depth estimation (MDE) even at low-level illumination. Therefore, existing complementary learning approaches for MDE fuse intensity information from images and scene details from event data for better scene understanding. However, most methods directly fuse two modalities at pixel level, ignoring that the attractive complementarity mainly impacts high-level patterns that only occupy a few pixels. For example, event data is likely to complement contours of scene objects. In this paper, we discretize the scene into a set of high-level patterns to explore the complementarity and propose a Pattern-based Complementary learning architecture for monocular Depth estimation (PCDepth). Concretely, PCDepth comprises two primary components: a complementary visual representation learning module for discretizing the scene into high-level patterns and integrating complementary patterns across modalities and a refined depth estimator aimed at scene reconstruction and depth prediction while maintaining an efficiency-accuracy balance. Through pattern-based complementary learning, PCDepth fully exploits two modalities and achieves more accurate predictions than existing methods, especially in challenging nighttime scenarios. Extensive experiments on MVSEC and DSEC datasets verify the effectiveness and superiority of our PCDepth. Remarkably, compared with state-of-the-art, PCDepth achieves a 37.9% improvement in accuracy in MVSEC nighttime scenarios.

Boyuan Zheng, Shouyi Lu, Renbo Huang et al.

We introduce R2LDM, an innovative approach for generating dense and accurate 4D radar point clouds, guided by corresponding LiDAR point clouds. Instead of utilizing range images or bird's eye view (BEV) images, we represent both LiDAR and 4D radar point clouds using voxel features, which more effectively capture 3D shape information. Subsequently, we propose the Latent Voxel Diffusion Model (LVDM), which performs the diffusion process in the latent space. Additionally, a novel Latent Point Cloud Reconstruction (LPCR) module is utilized to reconstruct point clouds from high-dimensional latent voxel features. As a result, R2LDM effectively generates LiDAR-like point clouds from paired raw radar data. We evaluate our approach on two different datasets, and the experimental results demonstrate that our model achieves 6- to 10-fold densification of radar point clouds, outperforming state-of-the-art baselines in 4D radar point cloud super-resolution. Furthermore, the enhanced radar point clouds generated by our method significantly improve downstream tasks, achieving up to 31.7% improvement in point cloud registration recall rate and 24.9% improvement in object detection accuracy.

Fan Lu, Quan Qi, Huaibin Qin

In the rapidly advancing information era, various human behaviors are being precisely recorded in the form of data, including identity information, criminal records, and communication data. Law enforcement agencies can effectively maintain social security and precisely combat criminal activities by analyzing the aforementioned data. In comparison to traditional data analysis methods, deep learning models, relying on the robust computational power in cloud centers, exhibit higher accuracy in extracting data features and inferring data. However, within the architecture of cloud centers, the transmission of data from end devices introduces significant latency, hindering real-time inference of data. Furthermore, low-latency edge computing architectures face limitations in direct deployment due to relatively weak computing and storage capacities of nodes. To address these challenges, a lightweight distributed knowledge graph completion architecture is proposed. Firstly, we introduce a lightweight distributed knowledge graph completion architecture that utilizes knowledge graph embedding for data analysis. Subsequently, to filter out substandard data, a personnel data quality assessment method named PDQA is proposed. Lastly, we present a model pruning algorithm that significantly reduces the model size while maximizing performance, enabling lightweight deployment. In experiments, we compare the effects of 11 advanced models on completing the knowledge graph of public security personnel information. The results indicate that the RotatE model outperforms other models significantly in knowledge graph completion, with the pruned model size reduced by 70\%, and hits@10 reaching 86.97\%.}

Jianyi Peng, Fan Lu, Bin Li et al.

Image-to-point cloud cross-modal Visual Place Recognition (VPR) is a challenging task where the query is an RGB image, and the database samples are LiDAR point clouds. Compared to single-modal VPR, this approach benefits from the widespread availability of RGB cameras and the robustness of point clouds in providing accurate spatial geometry and distance information. However, current methods rely on intermediate modalities that capture either the vertical or horizontal field of view, limiting their ability to fully exploit the complementary information from both sensors. In this work, we propose an innovative initial retrieval + re-rank method that effectively combines information from range (or RGB) images and Bird's Eye View (BEV) images. Our approach relies solely on a computationally efficient global descriptor similarity search process to achieve re-ranking. Additionally, we introduce a novel similarity label supervision technique to maximize the utility of limited training data. Specifically, we employ points average distance to approximate appearance similarity and incorporate an adaptive margin, based on similarity differences, into the vanilla triplet loss. Experimental results on the KITTI dataset demonstrate that our method significantly outperforms state-of-the-art approaches.

Yunnan Wang, Fan Lu, Kecheng Zheng et al.

Multimodal large language models (MLLMs) integrate image features from visual encoders with LLMs, demonstrating advanced comprehension capabilities. However, mainstream MLLMs are solely supervised by the next-token prediction of textual tokens, neglecting critical vision-centric information essential for analytical abilities. To track this dilemma, we introduce VaCo, which optimizes MLLM representations through Vision-Centric activation and Coordination from multiple vision foundation models (VFMs). VaCo introduces visual discriminative alignment to integrate task-aware perceptual features extracted from VFMs, thereby unifying the optimization of both textual and visual outputs in MLLMs. Specifically, we incorporate the learnable Modular Task Queries (MTQs) and Visual Alignment Layers (VALs) into MLLMs, activating specific visual signals under the supervision of diverse VFMs. To coordinate representation conflicts across VFMs, the crafted Token Gateway Mask (TGM) restricts the information flow among multiple groups of MTQs. Extensive experiments demonstrate that VaCo significantly improves the performance of different MLLMs on various benchmarks, showcasing its superior capabilities in visual comprehension.

Tianhang Wang, Fan Lu, Sanqing Qu et al.

Existing neural rendering-based urban scene reconstruction methods mainly focus on the Interpolated View Synthesis (IVS) setting that synthesizes from views close to training camera trajectory. However, IVS can not guarantee the on-par performance of the novel view outside the training camera distribution (\textit{e.g.}, looking left, right, or downwards), which limits the generalizability of the urban reconstruction application. Previous methods have optimized it via image diffusion, but they fail to handle text-ambiguous or large unseen view angles due to coarse-grained control of text-only diffusion. In this paper, we design UrbanCraft, which surmounts the Extrapolated View Synthesis (EVS) problem using hierarchical sem-geometric representations serving as additional priors. Specifically, we leverage the partially observable scene to reconstruct coarse semantic and geometric primitives, establishing a coarse scene-level prior through an occupancy grid as the base representation. Additionally, we incorporate fine instance-level priors from 3D bounding boxes to enhance object-level details and spatial relationships. Building on this, we propose the \textbf{H}ierarchical \textbf{S}emantic-Geometric-\textbf{G}uided Variational Score Distillation (HSG-VSD), which integrates semantic and geometric constraints from pretrained UrbanCraft2D into the score distillation sampling process, forcing the distribution to be consistent with the observable scene. Qualitative and quantitative comparisons demonstrate the effectiveness of our methods on EVS problem.

Zhiyuan Chen, Fan Lu, Guo Yu et al.

Tracking the 6DoF pose of unknown objects in monocular RGB video sequences is crucial for robotic manipulation. However, existing approaches typically rely on accurate depth information, which is non-trivial to obtain in real-world scenarios. Although depth estimation algorithms can be employed, geometric inaccuracy can lead to failures in RGBD-based pose tracking methods. To address this challenge, we introduce GSGTrack, a novel RGB-based pose tracking framework that jointly optimizes geometry and pose. Specifically, we adopt 3D Gaussian Splatting to create an optimizable 3D representation, which is learned simultaneously with a graph-based geometry optimization to capture the object's appearance features and refine its geometry. However, the joint optimization process is susceptible to perturbations from noisy pose and geometry data. Thus, we propose an object silhouette loss to address the issue of pixel-wise loss being overly sensitive to pose noise during tracking. To mitigate the geometric ambiguities caused by inaccurate depth information, we propose a geometry-consistent image pair selection strategy, which filters out low-confidence pairs and ensures robust geometric optimization. Extensive experiments on the OnePose and HO3D datasets demonstrate the effectiveness of GSGTrack in both 6DoF pose tracking and object reconstruction.

Fan Lu, Quan Qi, Huaibin Qin

Medical knowledge extraction methods based on edge computing deploy deep learning models on edge devices to achieve localized entity and relation extraction. This approach avoids transferring substantial sensitive data to cloud data centers, effectively safeguarding the privacy of healthcare services. However, existing relation extraction methods mainly employ a sequential pipeline approach, which classifies relations between determined entities after entity recognition. This mode faces challenges such as error propagation between tasks, insufficient consideration of dependencies between the two subtasks, and the neglect of interrelations between different relations within a sentence. To address these challenges, a joint extraction model with parameter sharing in edge computing is proposed, named CoEx-Bert. This model leverages shared parameterization between two models to jointly extract entities and relations. Specifically, CoEx-Bert employs two models, each separately sharing hidden layer parameters, and combines these two loss functions for joint backpropagation to optimize the model parameters. Additionally, it effectively resolves the issue of entity overlapping when extracting knowledge from unstructured Uyghur medical texts by considering contextual relations. Finally, this model is deployed on edge devices for real-time extraction and inference of Uyghur medical knowledge. Experimental results demonstrate that CoEx-Bert outperforms existing state-of-the-art methods, achieving accuracy, recall, and F1 scores of 90.65\%, 92.45\%, and 91.54\%, respectively, in the Uyghur traditional medical literature dataset. These improvements represent a 6.45\% increase in accuracy, a 9.45\% increase in recall, and a 7.95\% increase in F1 score compared to the baseline.

Fan Lu, Guang Chen, Yinlong Liu et al.

Point cloud registration is a fundamental problem in 3D computer vision. Outdoor LiDAR point clouds are typically large-scale and complexly distributed, which makes the registration challenging. In this paper, we propose an efficient hierarchical network named HRegNet for large-scale outdoor LiDAR point cloud registration. Instead of using all points in the point clouds, HRegNet performs registration on hierarchically extracted keypoints and descriptors. The overall framework combines the reliable features in deeper layer and the precise position information in shallower layers to achieve robust and precise registration. We present a correspondence network to generate correct and accurate keypoints correspondences. Moreover, bilateral consensus and neighborhood consensus are introduced for keypoints matching and novel similarity features are designed to incorporate them into the correspondence network, which significantly improves the registration performance. Besides, the whole network is also highly efficient since only a small number of keypoints are used for registration. Extensive experiments are conducted on two large-scale outdoor LiDAR point cloud datasets to demonstrate the high accuracy and efficiency of the proposed HRegNet. The project website is https://ispc-group.github.io/hregnet.

Fan Lu, Guang Chen, Yinlong Liu et al.

Predicting the future can significantly improve the safety of intelligent vehicles, which is a key component in autonomous driving. 3D point clouds accurately model 3D information of surrounding environment and are crucial for intelligent vehicles to perceive the scene. Therefore, prediction of 3D point clouds has great significance for intelligent vehicles, which can be utilized for numerous further applications. However, due to point clouds are unordered and unstructured, point cloud prediction is challenging and has not been deeply explored in current literature. In this paper, we propose a novel motion-based neural network named MoNet. The key idea of the proposed MoNet is to integrate motion features between two consecutive point clouds into the prediction pipeline. The introduction of motion features enables the model to more accurately capture the variations of motion information across frames and thus make better predictions for future motion. In addition, content features are introduced to model the spatial content of individual point clouds. A recurrent neural network named MotionRNN is proposed to capture the temporal correlations of both features. Besides, we propose an attention-based motion align module to address the problem of missing motion features in the inference pipeline. Extensive experiments on two large scale outdoor LiDAR datasets demonstrate the performance of the proposed MoNet. Moreover, we perform experiments on applications using the predicted point clouds and the results indicate the great application potential of the proposed method.

Sanqing Qu, Guang Chen, Dan Xu et al.

Online action detection is a task with the aim of identifying ongoing actions from streaming videos without any side information or access to future frames. Recent methods proposed to aggregate fixed temporal ranges of invisible but anticipated future frames representations as supplementary features and achieved promising performance. They are based on the observation that human beings often detect ongoing actions by contemplating the future vision simultaneously. However, we observed that at different action progressions, the optimal supplementary features should be obtained from distinct temporal ranges instead of simply fixed future temporal ranges. To this end, we introduce an adaptive features sampling strategy to overcome the mentioned variable-ranges of optimal supplementary features. Specifically, in this paper, we propose a novel Learning Action Progression Network termed LAP-Net, which integrates an adaptive features sampling strategy. At each time step, this sampling strategy first estimates current action progression and then decide what temporal ranges should be used to aggregate the optimal supplementary features. We evaluated our LAP-Net on three benchmark datasets, TVSeries, THUMOS-14 and HDD. The extensive experiments demonstrate that with our adaptive feature sampling strategy, the proposed LAP-Net can significantly outperform current state-of-the-art methods with a large margin.

Di Wu, Liang Ding, Fan Lu et al.

Slot filling and intent detection are two main tasks in spoken language understanding (SLU) system. In this paper, we propose a novel non-autoregressive model named SlotRefine for joint intent detection and slot filling. Besides, we design a novel two-pass iteration mechanism to handle the uncoordinated slots problem caused by conditional independence of non-autoregressive model. Experiments demonstrate that our model significantly outperforms previous models in slot filling task, while considerably speeding up the decoding (up to X 10.77). In-depth analyses show that 1) pretraining schemes could further enhance our model; 2) two-pass mechanism indeed remedy the uncoordinated slots.

Yanlin Zhou, Fan Lu, George Pu et al.

We propose a deep reinforcement learning (DRL) methodology for the tracking, obstacle avoidance, and formation control of nonholonomic robots. By separating vision-based control into a perception module and a controller module, we can train a DRL agent without sophisticated physics or 3D modeling. In addition, the modular framework averts daunting retrains of an image-to-action end-to-end neural network, and provides flexibility in transferring the controller to different robots. First, we train a convolutional neural network (CNN) to accurately localize in an indoor setting with dynamic foreground/background. Then, we design a new DRL algorithm named Momentum Policy Gradient (MPG) for continuous control tasks and prove its convergence. We also show that MPG is robust at tracking varying leader movements and can naturally be extended to problems of formation control. Leveraging reward shaping, features such as collision and obstacle avoidance can be easily integrated into a DRL controller.

Shuhang Chen, Adithya M. Devraj, Fan Lu et al.

Zap Q-learning is a recent class of reinforcement learning algorithms, motivated primarily as a means to accelerate convergence. Stability theory has been absent outside of two restrictive classes: the tabular setting, and optimal stopping. This paper introduces a new framework for analysis of a more general class of recursive algorithms known as stochastic approximation. Based on this general theory, it is shown that Zap Q-learning is consistent under a non-degeneracy assumption, even when the function approximation architecture is nonlinear. Zap Q-learning with neural network function approximation emerges as a special case, and is tested on examples from OpenAI Gym. Based on multiple experiments with a range of neural network sizes, it is found that the new algorithms converge quickly and are robust to choice of function approximation architecture.

Kong Aik Lee, Ville Hautamaki, Tomi Kinnunen et al.

The I4U consortium was established to facilitate a joint entry to NIST speaker recognition evaluations (SRE). The latest edition of such joint submission was in SRE 2018, in which the I4U submission was among the best-performing systems. SRE'18 also marks the 10-year anniversary of I4U consortium into NIST SRE series of evaluation. The primary objective of the current paper is to summarize the results and lessons learned based on the twelve sub-systems and their fusion submitted to SRE'18. It is also our intention to present a shared view on the advancements, progresses, and major paradigm shifts that we have witnessed as an SRE participant in the past decade from SRE'08 to SRE'18. In this regard, we have seen, among others, a paradigm shift from supervector representation to deep speaker embedding, and a switch of research challenge from channel compensation to domain adaptation.

Sanqing Qu, Guang Chen, Canbo Ye et al.

Detecting vehicles with strong robustness and high efficiency has become one of the key capabilities of fully autonomous driving cars. This topic has already been widely studied by GPU-accelerated deep learning approaches using image sensors and 3D LiDAR, however, few studies seek to address it with a horizontally mounted 2D laser scanner. 2D laser scanner is equipped on almost every autonomous vehicle for its superiorities in the field of view, lighting invariance, high accuracy and relatively low price. In this paper, we propose a highly efficient search-based L-Shape fitting algorithm for detecting positions and orientations of vehicles with a 2D laser scanner. Differing from the approach to formulating LShape fitting as a complex optimization problem, our method decomposes the L-Shape fitting into two steps: L-Shape vertexes searching and L-Shape corner localization. Our approach is computationally efficient due to its minimized complexity. In on-road experiments, our approach is capable of adapting to various circumstances with high efficiency and robustness.