Karthik Ramani

Dong In Lee, Hyungjun Doh, Seunggeun Chi et al.

Recent progress in 4D representations, such as Dynamic NeRF and 4D Gaussian Splatting (4DGS), has enabled dynamic 4D scene reconstruction. However, text-driven 4D scene editing remains under-explored due to the challenge of ensuring both multi-view and temporal consistency across space and time during editing. Existing studies rely on 2D diffusion models that edit frames independently, often causing motion distortion, geometric drift, and incomplete editing. We introduce Dynamic-eDiTor, a training-free text-driven 4D editing framework leveraging Multimodal Diffusion Transformer (MM-DiT) and 4DGS. This mechanism consists of Spatio-Temporal Sub-Grid Attention (STGA) for locally consistent cross-view and temporal fusion, and Context Token Propagation (CTP) for global propagation via token inheritance and optical-flow-guided token replacement. Together, these components allow Dynamic-eDiTor to perform seamless, globally consistent multi-view video without additional training and directly optimize pre-trained source 4DGS. Extensive experiments on multi-view video dataset DyNeRF demonstrate that our method achieves superior editing fidelity and both multi-view and temporal consistency prior approaches. Project page for results and code: https://di-lee.github.io/dynamic-eDiTor/

Wonseok Roh, Hwanhee Jung, Jong Wook Kim et al.

Recently, generalizable feed-forward methods based on 3D Gaussian Splatting have gained significant attention for their potential to reconstruct 3D scenes using finite resources. These approaches create a 3D radiance field, parameterized by per-pixel 3D Gaussian primitives, from just a few images in a single forward pass. However, unlike multi-view methods that benefit from cross-view correspondences, 3D scene reconstruction with a single-view image remains an underexplored area. In this work, we introduce CATSplat, a novel generalizable transformer-based framework designed to break through the inherent constraints in monocular settings. First, we propose leveraging textual guidance from a visual-language model to complement insufficient information from a single image. By incorporating scene-specific contextual details from text embeddings through cross-attention, we pave the way for context-aware 3D scene reconstruction beyond relying solely on visual cues. Moreover, we advocate utilizing spatial guidance from 3D point features toward comprehensive geometric understanding under single-view settings. With 3D priors, image features can capture rich structural insights for predicting 3D Gaussians without multi-view techniques. Extensive experiments on large-scale datasets demonstrate the state-of-the-art performance of CATSplat in single-view 3D scene reconstruction with high-quality novel view synthesis.

Mayank Patel, Rahul Jain, Asim Unmesh et al.

Understanding the motion of articulated mechanical assemblies from static geometry remains a core challenge in 3D perception and design automation. Prior work on everyday articulated objects such as doors and laptops typically assumes simplified kinematic structures or relies on joint annotations. However, in mechanical assemblies like gears, motion arises from geometric coupling, through meshing teeth or aligned axes, making it difficult for existing methods to reason about relational motion from geometry alone. To address this gap, we introduce MechBench, a benchmark dataset of 693 diverse synthetic gear assemblies with part-wise ground-truth motion trajectories. MechBench provides a structured setting to study coupled motion, where part dynamics are induced by contact and transmission rather than predefined joints. Building on this, we propose DYNAMO, a dependency-aware neural model that predicts per-part SE(3) motion trajectories directly from segmented CAD point clouds. Experiments show that DYNAMO outperforms strong baselines, achieving accurate and temporally consistent predictions across varied gear configurations. Together, MechBench and DYNAMO establish a novel systematic framework for data-driven learning of coupled mechanical motion in CAD assemblies.

Sangpil Kim, Jihyun Bae, Hyunggun Chi et al.

We introduce a multi-stage framework that uses mean curvature on a hand surface and focuses on learning interaction between hand and object by analyzing hand grasp type for hand action recognition in egocentric videos. The proposed method does not require 3D information of objects including 6D object poses which are difficult to annotate for learning an object's behavior while it interacts with hands. Instead, the framework synthesizes the mean curvature of the hand mesh model to encode the hand surface geometry in 3D space. Additionally, our method learns the hand grasp type which is highly correlated with the hand action. From our experiment, we notice that using hand grasp type and mean curvature of hand increases the performance of the hand action recognition.

Zongyue Zhao, Min Liu, Karthik Ramani

Traditional grid/neighbor-based static pooling has become a constraint for point cloud geometry analysis. In this paper, we propose DAR-Net, a novel network architecture that focuses on dynamic feature aggregation. The central idea of DAR-Net is generating a self-adaptive pooling skeleton that considers both scene complexity and local geometry features. Providing variable semi-local receptive fields and weights, the skeleton serves as a bridge that connect local convolutional feature extractors and a global recurrent feature integrator. Experimental results on indoor scene datasets show advantages of the proposed approach compared to state-of-the-art architectures that adopt static pooling methods.

Sangpil Kim, Nick Winovich, Guang Lin et al.

We propose a fully-convolutional conditional generative model, the latent transformation neural network (LTNN), capable of view synthesis using a light-weight neural network suited for real-time applications. In contrast to existing conditional generative models which incorporate conditioning information via concatenation, we introduce a dedicated network component, the conditional transformation unit (CTU), designed to learn the latent space transformations corresponding to specified target views. In addition, a consistency loss term is defined to guide the network toward learning the desired latent space mappings, a task-divided decoder is constructed to refine the quality of generated views, and an adaptive discriminator is introduced to improve the adversarial training process. The generality of the proposed methodology is demonstrated on a collection of three diverse tasks: multi-view reconstruction on real hand depth images, view synthesis of real and synthetic faces, and the rotation of rigid objects. The proposed model is shown to exceed state-of-the-art results in each category while simultaneously achieving a reduction in the computational demand required for inference by 30% on average.

Zhangjie Cao, Qixing Huang, Karthik Ramani

In this paper, we introduce a new method for classifying 3D objects. Our main idea is to project a 3D object onto a spherical domain centered around its barycenter and develop neural network to classify the spherical projection. We introduce two complementary projections. The first captures depth variations of a 3D object, and the second captures contour-information viewed from different angles. Spherical projections combine key advantages of two main-stream 3D classification methods: image-based and 3D-based. Specifically, spherical projections are locally planar, allowing us to use massive image datasets (e.g, ImageNet) for pre-training. Also spherical projections are similar to voxel-based methods, as they encode complete information of a 3D object in a single neural network capturing dependencies across different views. Our novel network design can fully utilize these advantages. Experimental results on ModelNet40 and ShapeNetCore show that our method is superior to prior methods.

Ayan Sinha, Asim Unmesh, Qixing Huang et al.

3D shape models are naturally parameterized using vertices and faces, \ie, composed of polygons forming a surface. However, current 3D learning paradigms for predictive and generative tasks using convolutional neural networks focus on a voxelized representation of the object. Lifting convolution operators from the traditional 2D to 3D results in high computational overhead with little additional benefit as most of the geometry information is contained on the surface boundary. Here we study the problem of directly generating the 3D shape surface of rigid and non-rigid shapes using deep convolutional neural networks. We develop a procedure to create consistent `geometry images' representing the shape surface of a category of 3D objects. We then use this consistent representation for category-specific shape surface generation from a parametric representation or an image by developing novel extensions of deep residual networks for the task of geometry image generation. Our experiments indicate that our network learns a meaningful representation of shape surfaces allowing it to interpolate between shape orientations and poses, invent new shape surfaces and reconstruct 3D shape surfaces from previously unseen images.