Hanlin Chen

Hanlin Chen, Jiaxin Wei, Xibin Song et al.

Frame-wise action-controlled image-to-video generation is a promising paradigm for interactive world simulation, where each control signal should elicit an immediate visual response. However, maintaining visual fidelity and 3D consistency over long autoregressive rollouts remains challenging. Existing 3D-aware methods often suffer from catastrophic drift due to two impediments: information loss from \textit{Latent--RGB Cycling}, where generated latents are repeatedly decoded to RGB and re-encoded for future conditioning, and the training--inference gap induced by the \textit{error-free hypothesis}, where clean training memory fails to match prediction-corrupted inference memory. To address these challenges, we present \textbf{Robust Dreamer}, a memory-augmented framework built around how to design 3D memory and how to use it robustly. First, we introduce \textbf{Latent Gaussian Memory}, which anchors diffusion latents inherited from the generation process to Gaussian primitives and recalls them via latent-space Gaussian splatting. This provides dense, geometry-aware, view-aligned conditioning while avoiding accumulated degradation from repeated VAE conversion. Second, we propose \textbf{Deviation Learning with Dynamic Deviation Archive}, which synthesizes rollout-induced latent deviations through a one-step approximation, stores them by autoregressive stage and denoising timestamp, and injects them into historical memory during training. This exposes the generator to realistic corrupted memory states and teaches internal correction before inference. Experiments on ScanNet, DL3DV, and OmniWorldGame demonstrate state-of-the-art long-horizon performance.

Yu Yang, Yue Liao, Jianbiao Mei et al.

Long-horizon action-conditioned video generation aims to synthesize temporally coherent videos that follow complex action instructions over extended horizons, requiring procedural ordering, persistent action execution, and scene consistency beyond conventional TI2V's short-term fidelity. Existing single-shot video generation models typically operate in an open-loop manner, leading to incomplete action execution, hallucinated motions, and temporal drift. To address this, we propose SPIRAL, a closed-loop framework that performs sequential planning and iterative reflection for action-conditioned long-horizon video generation. Specifically, SPIRAL instantiates a think-act-reflect process: a PlanAgent decomposes high-level goals into sub-actions, which condition a VideoGenerator to synthesize each segment alongside a memory context, while a CriticAgent evaluates intermediate video segments to provide corrective feedback for iterative refinement. This closed-loop design further supports self-evolution by utilizing PlanAgent-proposed actions and CriticAgent-derived rewards for GRPO-based post-training to enhance the video generator's long-horizon consistency. Moreover, we introduce ActVideoGen-Dataset for task-specific training, and establish ActVideoGen-Bench as a dedicated evaluation suite for measuring action quality and temporal coherence. Experiments across multiple TI2V backbones alongside the self-evolving strategy show consistent gains on ActVideoGen-Bench and VBench, demonstrating the effectiveness of SPIRAL.

Hanlin Chen, Chen Li, Mengqi Guo et al.

3D scene segmentation based on neural implicit representation has emerged recently with the advantage of training only on 2D supervision. However, existing approaches still requires expensive per-scene optimization that prohibits generalization to novel scenes during inference. To circumvent this problem, we introduce a generalizable 3D segmentation framework based on implicit representation. Specifically, our framework takes in multi-view image features and semantic maps as the inputs instead of only spatial information to avoid overfitting to scene-specific geometric and semantic information. We propose a novel soft voting mechanism to aggregate the 2D semantic information from different views for each 3D point. In addition to the image features, view difference information is also encoded in our framework to predict the voting scores. Intuitively, this allows the semantic information from nearby views to contribute more compared to distant ones. Furthermore, a visibility module is also designed to detect and filter out detrimental information from occluded views. Due to the generalizability of our proposed method, we can synthesize semantic maps or conduct 3D semantic segmentation for novel scenes with solely 2D semantic supervision. Experimental results show that our approach achieves comparable performance with scene-specific approaches. More importantly, our approach can even outperform existing strong supervision-based approaches with only 2D annotations. Our source code is available at: https://github.com/HLinChen/GNeSF.

Hanlin Chen, Renyuan Luo, Yiheng Feng

Prevalent solutions for Connected and Autonomous vehicle (CAV) mapping include high definition map (HD map) or real-time Simultaneous Localization and Mapping (SLAM). Both methods only rely on vehicle itself (onboard sensors or embedded maps) and can not adapt well to temporarily changed drivable areas such as work zones. Navigating CAVs in such areas heavily relies on how the vehicle defines drivable areas based on perception information. Difficulties in improving perception accuracy and ensuring the correct interpretation of perception results are challenging to the vehicle in these situations. This paper presents a prototype that introduces crowdsourcing trajectories information into the mapping process to enhance CAV's understanding on the drivable area and traffic rules. A Gaussian Mixture Model (GMM) is applied to construct the temporarily changed drivable area and occupancy grid map (OGM) based on crowdsourcing trajectories. The proposed method is compared with SLAM without any human driving information. Our method has adapted well with the downstream path planning and vehicle control module, and the CAV did not violate driving rule, which a pure SLAM method did not achieve.

Mengqi Guo, Chen Li, Hanlin Chen et al.

Recent neural implicit representations (NIRs) have achieved great success in the tasks of 3D reconstruction and novel view synthesis. However, they require the images of a scene from different camera views to be available for one-time training. This is expensive especially for scenarios with large-scale scenes and limited data storage. In view of this, we explore the task of incremental learning for NIRs in this work. We design a student-teacher framework to mitigate the catastrophic forgetting problem. Specifically, we iterate the process of using the student as the teacher at the end of each time step and let the teacher guide the training of the student in the next step. As a result, the student network is able to learn new information from the streaming data and retain old knowledge from the teacher network simultaneously. Although intuitive, naively applying the student-teacher pipeline does not work well in our task. Not all information from the teacher network is helpful since it is only trained with the old data. To alleviate this problem, we further introduce a random inquirer and an uncertainty-based filter to filter useful information. Our proposed method is general and thus can be adapted to different implicit representations such as neural radiance field (NeRF) and neural surface field. Extensive experimental results for both 3D reconstruction and novel view synthesis demonstrate the effectiveness of our approach compared to different baselines.

Changbai Li, Haodong Zhu, Hanlin Chen et al.

While 3D Gaussian Splatting (3DGS) enables high-quality, real-time rendering for bounded scenes, its extension to large-scale urban environments gives rise to critical challenges in terms of geometric consistency, memory efficiency, and computational scalability. To address these issues, we present UrbanGS, a scalable reconstruction framework that effectively tackles these challenges for city-scale applications. First, we propose a Depth-Consistent D-Normal Regularization module. Unlike existing approaches that rely solely on monocular normal estimators, which can effectively update rotation parameters yet struggle to update position parameters, our method integrates D-Normal constraints with external depth supervision. This allows for comprehensive updates of all geometric parameters. By further incorporating an adaptive confidence weighting mechanism based on gradient consistency and inverse depth deviation, our approach significantly enhances multi-view depth alignment and geometric coherence, which effectively resolves the issue of geometric accuracy in complex large-scale scenes. To improve scalability, we introduce a Spatially Adaptive Gaussian Pruning (SAGP) strategy, which dynamically adjusts Gaussian density based on local geometric complexity and visibility to reduce redundancy. Additionally, a unified partitioning and view assignment scheme is designed to eliminate boundary artifacts and optimize computational load. Extensive experiments on multiple urban datasets demonstrate that UrbanGS achieves superior performance in rendering quality, geometric accuracy, and memory efficiency, providing a systematic solution for high-fidelity large-scale scene reconstruction.

Yunsong Wang, Hanlin Chen, Gim Hee Lee

Recent advancements in vision-language foundation models have significantly enhanced open-vocabulary 3D scene understanding. However, the generalizability of existing methods is constrained due to their framework designs and their reliance on 3D data. We address this limitation by introducing Generalizable Open-Vocabulary Neural Semantic Fields (GOV-NeSF), a novel approach offering a generalizable implicit representation of 3D scenes with open-vocabulary semantics. We aggregate the geometry-aware features using a cost volume, and propose a Multi-view Joint Fusion module to aggregate multi-view features through a cross-view attention mechanism, which effectively predicts view-specific blending weights for both colors and open-vocabulary features. Remarkably, our GOV-NeSF exhibits state-of-the-art performance in both 2D and 3D open-vocabulary semantic segmentation, eliminating the need for ground truth semantic labels or depth priors, and effectively generalize across scenes and datasets without fine-tuning.

Yunsong Wang, Tianxin Huang, Hanlin Chen et al.

Recently, the integration of the efficient feed-forward scheme into 3D Gaussian Splatting (3DGS) has been actively explored. However, most existing methods focus on sparse view reconstruction of small regions and cannot produce eligible whole-scene reconstruction results in terms of either quality or efficiency. In this paper, we propose FreeSplat++, which focuses on extending the generalizable 3DGS to become an alternative approach to large-scale indoor whole-scene reconstruction, which has the potential of significantly accelerating the reconstruction speed and improving the geometric accuracy. To facilitate whole-scene reconstruction, we initially propose the Low-cost Cross-View Aggregation framework to efficiently process extremely long input sequences. Subsequently, we introduce a carefully designed pixel-wise triplet fusion method to incrementally aggregate the overlapping 3D Gaussian primitives from multiple views, adaptively reducing their redundancy. Furthermore, we propose a weighted floater removal strategy that can effectively reduce floaters, which serves as an explicit depth fusion approach that is crucial in whole-scene reconstruction. After the feed-forward reconstruction of 3DGS primitives, we investigate a depth-regularized per-scene fine-tuning process. Leveraging the dense, multi-view consistent depth maps obtained during the feed-forward prediction phase for an extra constraint, we refine the entire scene's 3DGS primitive to enhance rendering quality while preserving geometric accuracy. Extensive experiments confirm that our FreeSplat++ significantly outperforms existing generalizable 3DGS methods, especially in whole-scene reconstructions. Compared to conventional per-scene optimized 3DGS approaches, our method with depth-regularized per-scene fine-tuning demonstrates substantial improvements in reconstruction accuracy and a notable reduction in training time.

Hanlin Chen, Fangyin Wei, Gim Hee Lee

Humans naturally interact with their 3D surroundings using language, and modeling 3D language fields for scene understanding and interaction has gained growing interest. This paper introduces ChatSplat, a system that constructs a 3D language field, enabling rich chat-based interaction within 3D space. Unlike existing methods that primarily use CLIP-derived language features focused solely on segmentation, ChatSplat facilitates interaction on three levels: objects, views, and the entire 3D scene. For view-level interaction, we designed an encoder that encodes the rendered feature map of each view into tokens, which are then processed by a large language model (LLM) for conversation. At the scene level, ChatSplat combines multi-view tokens, enabling interactions that consider the entire scene. For object-level interaction, ChatSplat uses a patch-wise language embedding, unlike LangSplat's pixel-wise language embedding that implicitly includes mask and embedding. Here, we explicitly decouple the language embedding into separate mask and feature map representations, allowing more flexible object-level interaction. To address the challenge of learning 3D Gaussians posed by the complex and diverse distribution of language embeddings used in the LLM, we introduce a learnable normalization technique to standardize these embeddings, facilitating effective learning. Extensive experimental results demonstrate that ChatSplat supports multi-level interactions -- object, view, and scene -- within 3D space, enhancing both understanding and engagement.

Changbai Li, Haodong Zhu, Hanlin Chen et al.

3D Gaussian Splatting (3DGS) has made significant strides in real-time 3D scene reconstruction, but faces memory scalability issues in high-resolution scenarios. To address this, we propose Hierarchical Gaussian Splatting (HRGS), a memory-efficient framework with hierarchical block-level optimization. First, we generate a global, coarse Gaussian representation from low-resolution data. Then, we partition the scene into multiple blocks, refining each block with high-resolution data. The partitioning involves two steps: Gaussian partitioning, where irregular scenes are normalized into a bounded cubic space with a uniform grid for task distribution, and training data partitioning, where only relevant observations are retained for each block. By guiding block refinement with the coarse Gaussian prior, we ensure seamless Gaussian fusion across adjacent blocks. To reduce computational demands, we introduce Importance-Driven Gaussian Pruning (IDGP), which computes importance scores for each Gaussian and removes those with minimal contribution, speeding up convergence and reducing memory usage. Additionally, we incorporate normal priors from a pretrained model to enhance surface reconstruction quality. Our method enables high-quality, high-resolution 3D scene reconstruction even under memory constraints. Extensive experiments on three benchmarks show that HRGS achieves state-of-the-art performance in high-resolution novel view synthesis (NVS) and surface reconstruction tasks.

Hanlin Chen, Simin Chen, Wenyu Li et al.

As a safety-critical cyber-physical system, cybersecurity and related safety issues for Autonomous Vehicles (AVs) have been important research topics for a while. Among all the modules on AVs, perception is one of the most accessible attack surfaces, as drivers and AVs have no control over the outside environment. Most current work targeting perception security for AVs focuses on perception correctness. In this work, we propose an impact analysis based on inference time attacks for autonomous vehicles. We demonstrate in a simulation system that such inference time attacks can also threaten the safety of both the ego vehicle and other traffic participants.

Haodong Zhu, Changbai Li, Yangyang Ren et al.

Current multi-view 3D reconstruction methods rely on accurate camera calibration and pose estimation, requiring complex and time-intensive pre-processing that hinders their practical deployment. To address this challenge, we introduce Surf3R, an end-to-end feedforward approach that reconstructs 3D surfaces from sparse views without estimating camera poses and completes an entire scene in under 10 seconds. Our method employs a multi-branch and multi-view decoding architecture in which multiple reference views jointly guide the reconstruction process. Through the proposed branch-wise processing, cross-view attention, and inter-branch fusion, the model effectively captures complementary geometric cues without requiring camera calibration. Moreover, we introduce a D-Normal regularizer based on an explicit 3D Gaussian representation for surface reconstruction. It couples surface normals with other geometric parameters to jointly optimize the 3D geometry, significantly improving 3D consistency and surface detail accuracy. Experimental results demonstrate that Surf3R achieves state-of-the-art performance on multiple surface reconstruction metrics on ScanNet++ and Replica datasets, exhibiting excellent generalization and efficiency.

Jinnan Chen, Chen Li, Jianfeng Zhang et al.

In this work, we tackle the task of learning 3D human Gaussians from a single image, focusing on recovering detailed appearance and geometry including unobserved regions. We introduce a single-view generalizable Human Gaussian Model (HGM), which employs a novel generate-then-refine pipeline with the guidance from human body prior and diffusion prior. Our approach uses a ControlNet to refine rendered back-view images from coarse predicted human Gaussians, then uses the refined image along with the input image to reconstruct refined human Gaussians. To mitigate the potential generation of unrealistic human poses and shapes, we incorporate human priors from the SMPL-X model as a dual branch, propagating image features from the SMPL-X volume to the image Gaussians using sparse convolution and attention mechanisms. Given that the initial SMPL-X estimation might be inaccurate, we gradually refine it with our HGM model. We validate our approach on several publicly available datasets. Our method surpasses previous methods in both novel view synthesis and surface reconstruction. Our approach also exhibits strong generalization for cross-dataset evaluation and in-the-wild images.

Hanlin Chen, Fangyin Wei, Chen Li et al.

Although 3D Gaussian Splatting has been widely studied because of its realistic and efficient novel-view synthesis, it is still challenging to extract a high-quality surface from the point-based representation. Previous works improve the surface by incorporating geometric priors from the off-the-shelf normal estimator. However, there are two main limitations: 1) Supervising normals rendered from 3D Gaussians effectively updates the rotation parameter but is less effective for other geometric parameters; 2) The inconsistency of predicted normal maps across multiple views may lead to severe reconstruction artifacts. In this paper, we propose a Depth-Normal regularizer that directly couples normal with other geometric parameters, leading to full updates of the geometric parameters from normal regularization. We further propose a confidence term to mitigate inconsistencies of normal predictions across multiple views. Moreover, we also introduce a densification and splitting strategy to regularize the size and distribution of 3D Gaussians for more accurate surface modeling. Compared with Gaussian-based baselines, experiments show that our approach obtains better reconstruction quality and maintains competitive appearance quality at faster training speed and 100+ FPS rendering.

Hanlin Chen, Li'an Zhuo, Baochang Zhang et al.

Traditional neural architecture search (NAS) has a significant impact in computer vision by automatically designing network architectures for various tasks. In this paper, binarized neural architecture search (BNAS), with a search space of binarized convolutions, is introduced to produce extremely compressed models to reduce huge computational cost on embedded devices for edge computing. The BNAS calculation is more challenging than NAS due to the learning inefficiency caused by optimization requirements and the huge architecture space, and the performance loss when handling the wild data in various computing applications. To address these issues, we introduce operation space reduction and channel sampling into BNAS to significantly reduce the cost of searching. This is accomplished through a performance-based strategy that is robust to wild data, which is further used to abandon less potential operations. Furthermore, we introduce the Upper Confidence Bound (UCB) to solve 1-bit BNAS. Two optimization methods for binarized neural networks are used to validate the effectiveness of our BNAS. Extensive experiments demonstrate that the proposed BNAS achieves a comparable performance to NAS on both CIFAR and ImageNet databases. An accuracy of $96.53\%$ vs. $97.22\%$ is achieved on the CIFAR-10 dataset, but with a significantly compressed model, and a $40\%$ faster search than the state-of-the-art PC-DARTS. On the wild face recognition task, our binarized models achieve a performance similar to their corresponding full-precision models.

Hanlin Chen, Baochang Zhang, Song Xue et al.

Deep convolutional neural networks (DCNNs) have dominated as the best performers in machine learning, but can be challenged by adversarial attacks. In this paper, we defend against adversarial attacks using neural architecture search (NAS) which is based on a comprehensive search of denoising blocks, weight-free operations, Gabor filters and convolutions. The resulting anti-bandit NAS (ABanditNAS) incorporates a new operation evaluation measure and search process based on the lower and upper confidence bounds (LCB and UCB). Unlike the conventional bandit algorithm using UCB for evaluation only, we use UCB to abandon arms for search efficiency and LCB for a fair competition between arms. Extensive experiments demonstrate that ABanditNAS is faster than other NAS methods, while achieving an $8.73\%$ improvement over prior arts on CIFAR-10 under PGD-$7$.

Li'an Zhuo, Baochang Zhang, Linlin Yang et al.

Conventional learning methods simplify the bilinear model by regarding two intrinsically coupled factors independently, which degrades the optimization procedure. One reason lies in the insufficient training due to the asynchronous gradient descent, which results in vanishing gradients for the coupled variables. In this paper, we introduce a Cogradient Descent algorithm (CoGD) to address the bilinear problem, based on a theoretical framework to coordinate the gradient of hidden variables via a projection function. We solve one variable by considering its coupling relationship with the other, leading to a synchronous gradient descent to facilitate the optimization procedure. Our algorithm is applied to solve problems with one variable under the sparsity constraint, which is widely used in the learning paradigm. We validate our CoGD considering an extensive set of applications including image reconstruction, inpainting, and network pruning. Experiments show that it improves the state-of-the-art by a significant margin.

Li'an Zhuo, Baochang Zhang, Hanlin Chen et al.

Neural architecture search (NAS) proves to be among the best approaches for many tasks by generating an application-adaptive neural architecture, which is still challenged by high computational cost and memory consumption. At the same time, 1-bit convolutional neural networks (CNNs) with binarized weights and activations show their potential for resource-limited embedded devices. One natural approach is to use 1-bit CNNs to reduce the computation and memory cost of NAS by taking advantage of the strengths of each in a unified framework. To this end, a Child-Parent (CP) model is introduced to a differentiable NAS to search the binarized architecture (Child) under the supervision of a full-precision model (Parent). In the search stage, the Child-Parent model uses an indicator generated by the child and parent model accuracy to evaluate the performance and abandon operations with less potential. In the training stage, a kernel-level CP loss is introduced to optimize the binarized network. Extensive experiments demonstrate that the proposed CP-NAS achieves a comparable accuracy with traditional NAS on both the CIFAR and ImageNet databases. It achieves the accuracy of $95.27\%$ on CIFAR-10, $64.3\%$ on ImageNet with binarized weights and activations, and a $30\%$ faster search than prior arts.

Hanlin Chen, Li'an Zhuo, Baochang Zhang et al.

Neural architecture search (NAS) can have a significant impact in computer vision by automatically designing optimal neural network architectures for various tasks. A variant, binarized neural architecture search (BNAS), with a search space of binarized convolutions, can produce extremely compressed models. Unfortunately, this area remains largely unexplored. BNAS is more challenging than NAS due to the learning inefficiency caused by optimization requirements and the huge architecture space. To address these issues, we introduce channel sampling and operation space reduction into a differentiable NAS to significantly reduce the cost of searching. This is accomplished through a performance-based strategy used to abandon less potential operations. Two optimization methods for binarized neural networks are used to validate the effectiveness of our BNAS. Extensive experiments demonstrate that the proposed BNAS achieves a performance comparable to NAS on both CIFAR and ImageNet databases. An accuracy of $96.53\%$ vs. $97.22\%$ is achieved on the CIFAR-10 dataset, but with a significantly compressed model, and a $40\%$ faster search than the state-of-the-art PC-DARTS.