arXiv subjects

Cheng-Hao Kuo

Publications and source records attributed to Cheng-Hao Kuo.

At least 19 recordsLinked to original sources

arXiv2026

Memory Tree Guided Key Frame Querying for Efficient 3D Question Answering

Answering questions accurately and efficiently in embodied scenarios presents significant challenges due to limited computational and memory resources for Vision Language Model (VLM) inference. Existing methods adopt visual search key frame retrieval method to select critical question-related key frames for VLM input. However, visual search methods are inefficient because they require visual search among thousands of video frames for each individual user query. In this work, we propose a memory tree guided key frame selection paradigm for efficient 3D question answering in embodied scenarios. Our method leverages a compact and reusable 3D scene representation, termed MemTree3D, which supports real-time online construction leveraging camera 6-DoF poses. MemTree3D captures multi-level 3D scene information, enabling a Large Language Model to efficiently query and retrieve question-relevant key frames through our scoring-based frame selection without reprocessing the entire video stream. On OpenEQA, our method improves the LLM-Match of GPT-4o by 17.4%, LLaVA-OneVision-7B by 5.8%, outperforms existing visual search methods. Our code is available at https://github.com/hsiangwei0903/MemTree3D

cs.CV↗

arXiv2026

Enhancing Part-Level Point Grounding for Any Open-Source MLLMs

Visual grounding aims to associate free-form textual queries with specific regions in an image. While recent Multimodal Large Language Models (MLLMs) have demonstrated promising capabilities in this domain, they primarily excel at object-level grounding and often struggle with part-level grounding-an essential requirement for fine-grained tasks such as robotic manipulation. In this work, we introduce a general approach that equips any open-source MLLMs with accurate 2D part-level point grounding, offering a more direct alternative to conventional grounding representations. Our method leverages the attention mechanisms inherently present in MLLMs. By synthesizing text-conditioned, grounding-aware queries within intermediate layers via the proposed Q-Synth Module, we capture target-relevant attention patterns and refine them with a lightweight Attention-to-Point Decoder, which converts these patterns into a point-centric heatmap for final prediction. Notably, all original MLLM parameters are frozen, ensuring full preservation of their pre-trained capabilities. Experiments show that our design consistently improves part-level grounding accuracy across datasets and can be seamlessly integrated into any open-source MLLMs.

cs.CV↗

arXiv2026

Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models

Recent work explores new opportunities at the intersection of vision-language-action models (VLAs) and geometric foundation models (GFMs) for 3D reconstruction, such as VGGT. While the resulting geometric VLAs often show improved performance, it remains unclear (i) if modern VLAs already have sufficient geometric understanding to start with, (ii) what is the best architecture to inject geometric understanding into a VLA, and (iii) what is the effect of other design choices that affect geometric VLAs. In this paper we provide a rigorous experimental analysis to shed light on these questions, for a specific choice of VLA (GR00T-N1.5) and GFM (VGGT). Our first contribution is to formalize prior work's intuition that current VLAs lack geometric understanding, by providing a rigorous analysis based on linear probing. The analysis quantifies, for the first time, the "geometric gap" between VLAs and GFMs. Our second contribution is to identify and compare different strategies to bridge GFMs with VLAs. We implement three different architectures, which differ in the way they inject geometry in the VLA, while keeping low-level implementation details as similar as possible, to ensure a fair comparison. Finally, we analyze the impact of non-architectural choices (e.g., training data, number of cameras, reconstruction quality) on the performance of the geometric VLAs.

cs.CV↗

arXiv2026

Revisiting Model Stitching In the Foundation Model Era

Model stitching, connecting early layers of one model (source) to later layers of another (target) via a light stitch layer, has served as a probe of representational compatibility. Prior work finds that models trained on the same dataset remain stitchable (negligible accuracy drop) despite different initializations or objectives. We revisit stitching for Vision Foundation Models (VFMs) that vary in objectives, data, and modality mix (e.g., CLIP, DINOv2, SigLIP 2) and ask: Are heterogeneous VFMs stitchable? We introduce a systematic protocol spanning the stitch points, stitch layer families, training losses, and downstream tasks. Three findings emerge. (1) Stitch layer training matters: conventional approaches that match the intermediate features at the stitch point or optimize the task loss end-to-end struggle to retain accuracy, especially at shallow stitch points. (2) With a simple feature-matching loss at the target model's penultimate layer, heterogeneous VFMs become reliably stitchable across vision tasks. (3) For deep stitch points, the stitched model can surpass either constituent model at only a small inference overhead (for the stitch layer). Building on these findings, we further propose the VFM Stitch Tree (VST), which shares early layers across VFMs while retaining their later layers, yielding a controllable accuracy-latency trade-off for multimodal LLMs that often leverage multiple VFMs. Taken together, our study elevates stitching from a diagnostic probe to a practical recipe for integrating complementary VFM strengths and pinpointing where their representations align or diverge.

cs.CV↗

arXiv2025

Explicit Memory through Online 3D Gaussian Splatting Improves Class-Agnostic Video Segmentation

Remembering where object segments were predicted in the past is useful for improving the accuracy and consistency of class-agnostic video segmentation algorithms. Existing video segmentation algorithms typically use either no object-level memory (e.g. FastSAM) or they use implicit memories in the form of recurrent neural network features (e.g. SAM2). In this paper, we augment both types of segmentation models using an explicit 3D memory and show that the resulting models have more accurate and consistent predictions. For this, we develop an online 3D Gaussian Splatting (3DGS) technique to store predicted object-level segments generated throughout the duration of a video. Based on this 3DGS representation, a set of fusion techniques are developed, named FastSAM-Splat and SAM2-Splat, that use the explicit 3DGS memory to improve their respective foundation models' predictions. Ablation experiments are used to validate the proposed techniques' design and hyperparameter settings. Results from both real-world and simulated benchmarking experiments show that models which use explicit 3D memories result in more accurate and consistent predictions than those which use no memory or only implicit neural network memories. Project Page: https://topipari.com/projects/FastSAM-Splat/

cs.RO↗

arXiv2025

Attribute-based Object Grounding and Robot Grasp Detection with Spatial Reasoning

Enabling robots to grasp objects specified through natural language is essential for effective human-robot interaction, yet it remains a significant challenge. Existing approaches often struggle with open-form language expressions and typically assume unambiguous target objects without duplicates. Moreover, they frequently rely on costly, dense pixel-wise annotations for both object grounding and grasp configuration. We present Attribute-based Object Grounding and Robotic Grasping (OGRG), a novel framework that interprets open-form language expressions and performs spatial reasoning to ground target objects and predict planar grasp poses, even in scenes containing duplicated object instances. We investigate OGRG in two settings: (1) Referring Grasp Synthesis (RGS) under pixel-wise full supervision, and (2) Referring Grasp Affordance (RGA) using weakly supervised learning with only single-pixel grasp annotations. Key contributions include a bi-directional vision-language fusion module and the integration of depth information to enhance geometric reasoning, improving both grounding and grasping performance. Experiment results show that OGRG outperforms strong baselines in tabletop scenes with diverse spatial language instructions. In RGS, it operates at 17.59 FPS on a single NVIDIA RTX 2080 Ti GPU, enabling potential use in closed-loop or multi-object sequential grasping, while delivering superior grounding and grasp prediction accuracy compared to all the baselines considered. Under the weakly supervised RGA setting, OGRG also surpasses baseline grasp-success rates in both simulation and real-robot trials, underscoring the effectiveness of its spatial reasoning design. Project page: https://z.umn.edu/ogrg

cs.RO↗

arXiv2025

OpenM3D: Open Vocabulary Multi-view Indoor 3D Object Detection without Human Annotations

Open-vocabulary (OV) 3D object detection is an emerging field, yet its exploration through image-based methods remains limited compared to 3D point cloud-based methods. We introduce OpenM3D, a novel open-vocabulary multi-view indoor 3D object detector trained without human annotations. In particular, OpenM3D is a single-stage detector adapting the 2D-induced voxel features from the ImGeoNet model. To support OV, it is jointly trained with a class-agnostic 3D localization loss requiring high-quality 3D pseudo boxes and a voxel-semantic alignment loss requiring diverse pre-trained CLIP features. We follow the training setting of OV-3DET where posed RGB-D images are given but no human annotations of 3D boxes or classes are available. We propose a 3D Pseudo Box Generation method using a graph embedding technique that combines 2D segments into coherent 3D structures. Our pseudo-boxes achieve higher precision and recall than other methods, including the method proposed in OV-3DET. We further sample diverse CLIP features from 2D segments associated with each coherent 3D structure to align with the corresponding voxel feature. The key to training a highly accurate single-stage detector requires both losses to be learned toward high-quality targets. At inference, OpenM3D, a highly efficient detector, requires only multi-view images for input and demonstrates superior accuracy and speed (0.3 sec. per scene) on ScanNet200 and ARKitScenes indoor benchmarks compared to existing methods. We outperform a strong two-stage method that leverages our class-agnostic detector with a ViT CLIP-based OV classifier and a baseline incorporating multi-view depth estimator on both accuracy and speed.

cs.CV↗

arXiv2025

Details Matter for Indoor Open-vocabulary 3D Instance Segmentation

Unlike closed-vocabulary 3D instance segmentation that is often trained end-to-end, open-vocabulary 3D instance segmentation (OV-3DIS) often leverages vision-language models (VLMs) to generate 3D instance proposals and classify them. While various concepts have been proposed from existing research, we observe that these individual concepts are not mutually exclusive but complementary. In this paper, we propose a new state-of-the-art solution for OV-3DIS by carefully designing a recipe to combine the concepts together and refining them to address key challenges. Our solution follows the two-stage scheme: 3D proposal generation and instance classification. We employ robust 3D tracking-based proposal aggregation to generate 3D proposals and remove overlapped or partial proposals by iterative merging/removal. For the classification stage, we replace the standard CLIP model with Alpha-CLIP, which incorporates object masks as an alpha channel to reduce background noise and obtain object-centric representation. Additionally, we introduce the standardized maximum similarity (SMS) score to normalize text-to-proposal similarity, effectively filtering out false positives and boosting precision. Our framework achieves state-of-the-art performance on ScanNet200 and S3DIS across all AP and AR metrics, even surpassing an end-to-end closed-vocabulary method.

cs.CV↗

arXiv2025

Advancing Multimodal LLMs by Large-Scale 3D Visual Instruction Dataset Generation

Multimodal Large Language Models (MLLMs) struggle with accurately capturing camera-object relations, especially for object orientation, camera viewpoint, and camera shots. This stems from the fact that existing MLLMs are trained on images with limited diverse camera-object relations and corresponding textual descriptions. To address this, we propose a synthetic generation pipeline to create large-scale 3D visual instruction datasets. Our framework takes 3D assets as input and uses rendering and diffusion-based image generation models to create photorealistic images preserving precise camera-object relations. Additionally, large language models (LLMs) are used to generate text prompts for guiding visual instruction tuning and controlling image generation. We create Ultimate3D, a dataset of 240K VQAs with precise camera-object annotations, and corresponding benchmark. MLLMs fine-tuned on our proposed dataset outperform commercial models by a large margin, achieving an average accuracy improvement of 33.4% on camera-object relation recognition tasks. Our code, dataset, and benchmark will contribute to broad MLLM applications.

cs.GR↗

arXiv2025

UA-Pose: Uncertainty-Aware 6D Object Pose Estimation and Online Object Completion with Partial References

6D object pose estimation has shown strong generalizability to novel objects. However, existing methods often require either a complete, well-reconstructed 3D model or numerous reference images that fully cover the object. Estimating 6D poses from partial references, which capture only fragments of an object's appearance and geometry, remains challenging. To address this, we propose UA-Pose, an uncertainty-aware approach for 6D object pose estimation and online object completion specifically designed for partial references. We assume access to either (1) a limited set of RGBD images with known poses or (2) a single 2D image. For the first case, we initialize a partial object 3D model based on the provided images and poses, while for the second, we use image-to-3D techniques to generate an initial object 3D model. Our method integrates uncertainty into the incomplete 3D model, distinguishing between seen and unseen regions. This uncertainty enables confidence assessment in pose estimation and guides an uncertainty-aware sampling strategy for online object completion, enhancing robustness in pose estimation accuracy and improving object completeness. We evaluate our method on the YCB-Video, YCBInEOAT, and HO3D datasets, including RGBD sequences of YCB objects manipulated by robots and human hands. Experimental results demonstrate significant performance improvements over existing methods, particularly when object observations are incomplete or partially captured. Project page: https://minfenli.github.io/UA-Pose/

cs.CV↗

arXiv2025

POp-GS: Next Best View in 3D-Gaussian Splatting with P-Optimality

In this paper, we present a novel algorithm for quantifying uncertainty and information gained within 3D Gaussian Splatting (3D-GS) through P-Optimality. While 3D-GS has proven to be a useful world model with high-quality rasterizations, it does not natively quantify uncertainty or information, posing a challenge for real-world applications such as 3D-GS SLAM. We propose to quantify information gain in 3D-GS by reformulating the problem through the lens of optimal experimental design, which is a classical solution widely used in literature. By restructuring information quantification of 3D-GS through optimal experimental design, we arrive at multiple solutions, of which T-Optimality and D-Optimality perform the best quantitatively and qualitatively as measured on two popular datasets. Additionally, we propose a block diagonal covariance approximation which provides a measure of correlation at the expense of a greater computation cost.

cs.CV↗

arXiv2025

CSCPR: Cross-Source-Context Indoor RGB-D Place Recognition

We extend our previous work, PoCo, and present a new algorithm, Cross-Source-Context Place Recognition (CSCPR), for RGB-D indoor place recognition that integrates global retrieval and reranking into an end-to-end model and keeps the consistency of using Context-of-Clusters (CoCs) for feature processing. Unlike prior approaches that primarily focus on the RGB domain for place recognition reranking, CSCPR is designed to handle the RGB-D data. We apply the CoCs to handle cross-sourced and cross-scaled RGB-D point clouds and introduce two novel modules for reranking: the Self-Context Cluster (SCC) and the Cross Source Context Cluster (CSCC), which enhance feature representation and match query-database pairs based on local features, respectively. We also release two new datasets, ScanNetIPR and ARKitIPR. Our experiments demonstrate that CSCPR significantly outperforms state-of-the-art models on these datasets by at least 29.27% in Recall@1 on the ScanNet-PR dataset and 43.24% in the new datasets. Code and datasets will be released.

cs.CV↗

arXiv2024

PoCo: Point Context Cluster for RGBD Indoor Place Recognition

We present a novel end-to-end algorithm (PoCo) for the indoor RGB-D place recognition task, aimed at identifying the most likely match for a given query frame within a reference database. The task presents inherent challenges attributed to the constrained field of view and limited range of perception sensors. We propose a new network architecture, which generalizes the recent Context of Clusters (CoCs) to extract global descriptors directly from the noisy point clouds through end-to-end learning. Moreover, we develop the architecture by integrating both color and geometric modalities into the point features to enhance the global descriptor representation. We conducted evaluations on public datasets ScanNet-PR and ARKit with 807 and 5047 scenarios, respectively. PoCo achieves SOTA performance: on ScanNet-PR, we achieve R@1 of 64.63%, a 5.7% improvement from the best-published result CGis (61.12%); on Arkit, we achieve R@1 of 45.12%, a 13.3% improvement from the best-published result CGis (39.82%). In addition, PoCo shows higher efficiency than CGis in inference time (1.75X-faster), and we demonstrate the effectiveness of PoCo in recognizing places within a real-world laboratory environment.

cs.CV↗

arXiv2024

V-MIND: Building Versatile Monocular Indoor 3D Detector with Diverse 2D Annotations

The field of indoor monocular 3D object detection is gaining significant attention, fueled by the increasing demand in VR/AR and robotic applications. However, its advancement is impeded by the limited availability and diversity of 3D training data, owing to the labor-intensive nature of 3D data collection and annotation processes. In this paper, we present V-MIND (Versatile Monocular INdoor Detector), which enhances the performance of indoor 3D detectors across a diverse set of object classes by harnessing publicly available large-scale 2D datasets. By leveraging well-established monocular depth estimation techniques and camera intrinsic predictors, we can generate 3D training data by converting large-scale 2D images into 3D point clouds and subsequently deriving pseudo 3D bounding boxes. To mitigate distance errors inherent in the converted point clouds, we introduce a novel 3D self-calibration loss for refining the pseudo 3D bounding boxes during training. Additionally, we propose a novel ambiguity loss to address the ambiguity that arises when introducing new classes from 2D datasets. Finally, through joint training with existing 3D datasets and pseudo 3D bounding boxes derived from 2D datasets, V-MIND achieves state-of-the-art object detection performance across a wide range of classes on the Omni3D indoor dataset.

cs.CV↗

arXiv2024

Enhancing Single Image to 3D Generation using Gaussian Splatting and Hybrid Diffusion Priors

3D object generation from a single image involves estimating the full 3D geometry and texture of unseen views from an unposed RGB image captured in the wild. Accurately reconstructing an object's complete 3D structure and texture has numerous applications in real-world scenarios, including robotic manipulation, grasping, 3D scene understanding, and AR/VR. Recent advancements in 3D object generation have introduced techniques that reconstruct an object's 3D shape and texture by optimizing the efficient representation of Gaussian Splatting, guided by pre-trained 2D or 3D diffusion models. However, a notable disparity exists between the training datasets of these models, leading to distinct differences in their outputs. While 2D models generate highly detailed visuals, they lack cross-view consistency in geometry and texture. In contrast, 3D models ensure consistency across different views but often result in overly smooth textures. We propose bridging the gap between 2D and 3D diffusion models to address this limitation by integrating a two-stage frequency-based distillation loss with Gaussian Splatting. Specifically, we leverage geometric priors in the low-frequency spectrum from a 3D diffusion model to maintain consistent geometry and use a 2D diffusion model to refine the fidelity and texture in the high-frequency spectrum of the generated 3D structure, resulting in more detailed and fine-grained outcomes. Our approach enhances geometric consistency and visual quality, outperforming the current SOTA. Additionally, we demonstrate the easy adaptability of our method for efficient object pose estimation and tracking.

cs.CV↗

arXiv2024

Modeling Uncertainty in 3D Gaussian Splatting through Continuous Semantic Splatting

In this paper, we present a novel algorithm for probabilistically updating and rasterizing semantic maps within 3D Gaussian Splatting (3D-GS). Although previous methods have introduced algorithms which learn to rasterize features in 3D-GS for enhanced scene understanding, 3D-GS can fail without warning which presents a challenge for safety-critical robotic applications. To address this gap, we propose a method which advances the literature of continuous semantic mapping from voxels to ellipsoids, combining the precise structure of 3D-GS with the ability to quantify uncertainty of probabilistic robotic maps. Given a set of images, our algorithm performs a probabilistic semantic update directly on the 3D ellipsoids to obtain an expectation and variance through the use of conjugate priors. We also propose a probabilistic rasterization which returns per-pixel segmentation predictions with quantifiable uncertainty. We compare our method with similar probabilistic voxel-based methods to verify our extension to 3D ellipsoids, and perform ablation studies on uncertainty quantification and temporal smoothing.

cs.RO↗

arXiv2024

Configurable Embodied Data Generation for Class-Agnostic RGB-D Video Segmentation

This paper presents a method for generating large-scale datasets to improve class-agnostic video segmentation across robots with different form factors. Specifically, we consider the question of whether video segmentation models trained on generic segmentation data could be more effective for particular robot platforms if robot embodiment is factored into the data generation process. To answer this question, a pipeline is formulated for using 3D reconstructions (e.g. from HM3DSem) to generate segmented videos that are configurable based on a robot's embodiment (e.g. sensor type, sensor placement, and illumination source). A resulting massive RGB-D video panoptic segmentation dataset (MVPd) is introduced for extensive benchmarking with foundation and video segmentation models, as well as to support embodiment-focused research in video segmentation. Our experimental findings demonstrate that using MVPd for finetuning can lead to performance improvements when transferring foundation models to certain robot embodiments, such as specific camera placements. These experiments also show that using 3D modalities (depth images and camera pose) can lead to improvements in video segmentation accuracy and consistency. The project webpage is available at https://topipari.com/projects/MVPd

cs.RO↗

arXiv2024

GenRC: Generative 3D Room Completion from Sparse Image Collections

Sparse RGBD scene completion is a challenging task especially when considering consistent textures and geometries throughout the entire scene. Different from existing solutions that rely on human-designed text prompts or predefined camera trajectories, we propose GenRC, an automated training-free pipeline to complete a room-scale 3D mesh with high-fidelity textures. To achieve this, we first project the sparse RGBD images to a highly incomplete 3D mesh. Instead of iteratively generating novel views to fill in the void, we utilized our proposed E-Diffusion to generate a view-consistent panoramic RGBD image which ensures global geometry and appearance consistency. Furthermore, we maintain the input-output scene stylistic consistency through textual inversion to replace human-designed text prompts. To bridge the domain gap among datasets, E-Diffusion leverages models trained on large-scale datasets to generate diverse appearances. GenRC outperforms state-of-the-art methods under most appearance and geometric metrics on ScanNet and ARKitScenes datasets, even though GenRC is not trained on these datasets nor using predefined camera trajectories. Project page: https://minfenli.github.io/GenRC

cs.CV↗