Millimeter-wave (mmWave) point-cloud radar shows great promise in enabling responsive human-machine interfaces (e.g., through pose and gesture tracking and for emerging augmented reality approaches). However, generating dense and temporally consistent 3D human point clouds from sequential mmWave signals is challenging due to point-cloud sparsity, jitter, and noise. Existing approaches have made progress in single-frame densification, but are inaccurate over multiple frames. This work redefines the problem as a 3D point cloud denoising task, leveraging reverse diffusion processes to transform sparse mmWave data into detailed and accurate whole-body representations. Our proposed method, mmDiffusion, effectively exploits diffusion models and temporal context within mmWave sequences to learn the denoising process, resulting in denser and temporally coherent human point clouds. For the first time, we also introduce an evaluation metric tailored to measure temporal consistency for sequential 3D human point clouds. Experimental results demonstrate that mmDiffusion significantly outperforms existing methods.

The rapid development of diffusion models has significantly advanced AI-generated content (AIGC), particularly in Text-to-Image (T2I) and Text-to-Video (T2V) generation. Text-based video editing, leveraging these generative capabilities, has emerged as a promising field, enabling precise modifications to videos based on text prompts. Despite the proliferation of innovative video editing models, there is a conspicuous lack of comprehensive evaluation benchmarks that holistically assess these models’ performance across various dimensions. Existing evaluations are limited and inconsistent, typically summarizing overall performance with a single score, which obscures models’ effectiveness on individual editing tasks. To address this gap, we propose EditBoard, the first comprehensive evaluation benchmark for text-based video editing models. EditBoard encompasses nine automatic metrics across four dimensions, evaluating models on four task categories and introducing three new metrics to assess fidelity. This task-oriented benchmark facilitates objective evaluation by detailing model performance and providing insights into each model’s strengths and weaknesses. By open-sourcing EditBoard, we aim to standardize evaluation and advance the development of robust video editing models.

The performance of 3D object detection in large outdoor point clouds deteriorates significantly in an unseen environment due to the inter-domain gap. To address these challenges, most existing methods for domain adaptation harness self-training schemes and attempt to bridge the gap by focusing on a single factor that causes the inter-domain gap, such as objects' sizes, shapes, and foreground density variation. However, the resulting adaptations suggest that there is still a substantial inter-domain gap left to be minimized. We argue that this is due to two limitations: 1) Biased pseudo-label collection from self-training. 2) Multiple factors jointly contributing to how the object is perceived in the unseen target domain. In this work, we propose a grouping-exploration strategy framework, Group Explorer Domain Adaptation ($\textbf{GroupEXP-DA}$), to addresses those two issues. Specifically, our grouping divides the available label sets into multiple clusters and ensures all of them have equal learning attention with the group-equivariant spatial feature, avoiding dominant types of objects causing imbalance problems. Moreover, grouping learns to divide objects by considering inherent factors in a data-driven manner, without considering each factor separately as existing works. On top of the group-equivariant spatial feature that selectively detects objects similar to the input group, we additionally introduce an explorative group update strategy that reduces the false negative detection in the target domain, further reducing the inter-domain gap. During inference, only the learned group features are necessary for making the group-equivariant spatial feature, placing our method as a simple add-on that can be applicable to most existing detectors. We show how each module contributes to substantially bridging the inter-domain gaps compared to existing works across large urban outdoor datasets such as NuScenes, Waymo, and KITTI.

Pedestrian detection is a fundamental task for many downstream applications. Visible and thermal images, as the two most important data types, are usually used to detect pedestrians under various environmental conditions. Many state-of-the-art works have been proposed to use two-stream (i.e., two-branch) architectures to combine visible and thermal information to improve detection performance. However, conventional visible-thermal fusion-based methods have no ability to obtain useful information from the visible branch under poor visibility conditions. The visible branch could even sometimes bring noise into the combined features. In this paper, we present a novel thermal and visible fusion architecture for pedestrian detection. Instead of simply using two branches to separately extract thermal and visible features and then fusing them, we introduce a hallucination branch to learn the mapping from the thermal to the visible domain, forming a novel three-branch feature extraction module. We then adaptively fuse feature maps from all three branches (i.e., thermal, visible, and hallucination). With this new integrated hallucination branch, our network can still get relatively good visible feature maps under challenging low-visibility conditions, thus boosting the overall detection performance. Finally, we experimentally demonstrate the superiority of the proposed architecture over conventional fusion methods.

Although the application of Transformers to 3-D point cloud processing has achieved significant progress and success, it is still challenging for existing 3-D Transformer methods to efficiently and accurately learn both valuable global and local features for improved applications. This article presents a novel point cloud representational learning network, called 3-D Dual Self-attention global local (GLocal) Transformer Network (3DGTN), for improved feature learning in both classification and segmentation tasks, with the following key contributions. First, a GLocal feature learning (GFL) block with the dual self-attention mechanism [i.e., a novel point-patch self-attention, called PPSA, and a channel-wise self-attention (CSA)] is designed to efficiently learn the global and local context information. Second, the GFL block is integrated with a multiscale Graph Convolution-based local feature aggregation (LFA) block, leading to a GLocal information extraction module that can efficiently capture critical information. Third, a series of GLocal modules are used to construct a new hierarchical encoder–decoder structure to enable the learning of information in different scales in a hierarchical manner. The proposed framework is evaluated on both classification and segmentation datasets, demonstrating that the proposed method is capable of outperforming many state-of-the-art methods on both synthetic and LiDAR data. Our code has been released athttps://github.com/d62lu/3DGTN.

Thermal imagery is emerging as a viable candidate for 24-7, all-weather pedestrian detection owning to thermal sensors’ robust performance for pedestrian detection under different weather and illumination conditions. Despite the promising results obtained from combining visible (RGB) and thermal cameras in multi-spectral fusion techniques, the complex synchronization requirements, including alignment and calibration of sensors, impede their deployment in real-world scenarios. In this paper, we introduce a novel approach for domain adaptation to enhance the performance of pedestrian detection based solely on thermal images. Our proposed approach involves several stages. Firstly, we use both thermal and visible images as input during the training phase. Secondly, we leverage a thermal-to-visible hallucination network to generate feature maps that are similar to those generated by the visible branch. Finally, we design a transformer-based multi-modal fusion module to integrate the hallucinated visible and thermal information more effectively. The thermal-to-visible hallucination network acts as domain adaptation, allowing us to obtain pseudo-visual and thermal features using solely thermal input. Based on the experimental results, it is observed the mean average precision (mAP) increases by 4.72% and the miss rate decreases by 7.56% on the KAIST dataset when compared to the baseline model.

Abstract The intricacy of 3D surfaces often results cutting‐edge point cloud denoising (PCD) models in surface degradation including remnant noise, wrongly‐removed geometric details. Although using multi‐scale patches to encode the geometry of a point has become the common wisdom in PCD, we find that simple aggregation of extracted multi‐scale features can not adaptively utilize the appropriate scale information according to the geometric information around noisy points. It leads to surface degradation, especially for points close to edges and points on complex curved surfaces. We raise an intriguing question – if employing multi‐scale geometric perception information to guide the network to utilize multi‐scale information, can eliminate the severe surface degradation problem? To answer it, we propose a Multi‐offset Denoising Network (MODNet) customized for multi‐scale patches. First, we extract the low‐level feature of three scales patches by patch feature encoders. Second, a multi‐scale perception module is designed to embed multi‐scale geometric information for each scale feature and regress multi‐scale weights to guide a multi‐offset denoising displacement. Third, a multi‐offset decoder regresses three scale offsets, which are guided by the multi‐scale weights to predict the final displacement by weighting them adaptively. Experiments demonstrate that our method achieves new state‐of‐the‐art performance on both synthetic and real‐scanned datasets. Our code is publicly available at https://github.com/hay-001/MODNet .

Recent 3-D scanning techniques can produce various kinds of digitized 3-D data. Most of these scanned data are in a format of unstructured point clouds. Such low-level representation of 3-D data usually contains only geometric properties (point positions), while lacking higher level structure cues, for example, feature lines. Feature lines can be defined as a visually prominent characteristic of the shape, including edges, ridges, and valley lines in multiple scales, which can support a lot of downstream applications, such as shape reconstruction and analysis. We present a two-phase algorithm for extracting line-type features on point clouds. To extract both large-scale and shallow feature lines, we first define a statistical metric to detect all potential feature points while immune to the noise to some extent. Then, for correctly reconstructing the feature lines from these identified coarse feature points, we introduce an anisotropic contracting scheme to force feature points lying on the underlying real feature lines. To illustrate the reliability of our method, various experiments have been conducted on both synthetic and raw data. Both visual and quantitative comparisons show that our method is robust to noise and can correctly extract multiscale feature lines. In addition, our method is generally applicable to robotic picking.Note to Practitioners—This article was motivated by the problem of the feature line extraction for real scanned point clouds. Feature lines, as one kind of the most important structure information, depict the basic shape of the real object in our life. Extracting this kind of shape features from the unstructured point clouds can facilitate a variety of downstream practical applications, such as product design, workpiece manufacturing, and robotic grasping. Existing approaches to detect features either heavily rely on differential quantities, which are sensitive to the noise, or need an elaborately designed local descriptor but fail to recognize small-scale features. These challenges motivate us to design a new approach aiming at extracting multiscale feature lines while keeping robustness to heavy noise. The technique developed in this work can produce high-quality feature points and feature lines, which would serve as higher level structural information and facilitate many applications. Additional applications in 6-degree-of-freedom (6-DoF) pose estimation demonstrate the potential of our method for robotic picking.

Point cloud classification is a fundamental task in 3D applications. However, it is challenging to achieve effective feature learning due to the irregularity and unordered nature of point clouds. Lately, 3D Transformers have been adopted to improve point cloud processing. Nevertheless, massive Transformer layers tend to incur huge computational and memory costs. This paper presented a novel hierarchical framework that incorporated convolutions with Transformers for point cloud classification, named 3D Convolution-Transformer Network (3DCTN). It combined the strong local feature learning ability of convolutions with the remarkable global context modeling capability of Transformers. Our method had two main modules operating on the downsampling point sets. Each module consisted of a multi-scale local feature aggregating (LFA) block and a global feature learning (GFL) block, which were implemented by using the Graph Convolution and Transformer respectively. We also conducted a detailed investigation on a series of self-attention variants to explore better performance for our network. Various experiments on ModelNet40 and ScanObjectNN datasets demonstrated that our method achieves state-of-the-art classification performance with a lightweight design. The code is publicly available athttps://github.com/d62lu/3DCTN.

In this article, we propose a Multi-feature Fusion VoteNet (MFFVoteNet) framework for improving the 3D object detection performance in cluttered and heavily occluded scenes. Our method takes the point cloud and the synchronized RGB image as inputs to provide object detection results in 3D space. Our detection architecture is built on VoteNet with three key designs. First, we augment the VoteNet input with point color information to enhance the difference of various instances in a scene. Next, we integrate an image feature module into the VoteNet to provide a strong object class signal that can facilitate deterministic detections in occlusion. Moreover, we propose a Projection Non-Maximum Suppression (PNMS) method in 3D object detection to eliminate redundant proposals and hence provide more accurate positioning of 3D objects. We evaluate the proposed MFFVoteNet on two challenging 3D object detection datasets, i.e., ScanNetv2 and SUN RGB-D. Extensive experiments show that our framework can effectively improve the performance of 3D object detection.

Benefiting from the excellent performance of Siamese-based trackers, huge progress on 2D visual tracking has been achieved. However, 3D visual tracking is still under-explored. Inspired by the idea of Hough voting in 3D object detection, in this paper, we propose a Multi-level Voting Siamese Network (MLVSNet) for 3D visual tracking from outdoor point cloud sequences. To deal with sparsity in outdoor 3D point clouds, we propose to perform Hough voting on multi-level features to get more vote centers and retain more useful information, instead of voting only on the fi-nal level feature as in previous methods. We also design an efficient and lightweight Target-Guided Attention (TGA) module to transfer the target information and highlight the target points in the search area. Moreover, we propose a Vote-cluster Feature Enhancement (VFE) module to exploit the relationships between different vote clusters. Extensive experiments on the 3D tracking benchmark of KITTI dataset demonstrate that our MLVSNet outperforms state-of-the-art methods with significant margins. Code will be available at https://github.com/CodeWZT/MLVSNet.

Hough voting, as has been demonstrated in VoteNet, is effective for 3D object detection, where voting is a key step. In this paper, we propose a novel VoteNet-based 3D detector with vote enhancement to improve the detection accuracy in cluttered indoor scenes. It addresses the limitations of current voting schemes, i.e., votes from neighboring objects and background have significant negative impacts. Before voting, we replace the classic MLP with the proposed Attentive MLP (AMLP) in the backbone network to get better feature description of seed points. During voting, we design a new vote attraction loss (VALoss) to enforce vote centers to locate closely and compactly to the corresponding object centers. After voting, we then devise a vote weighting module to integrate the foreground/background prediction into the vote aggregation process to enhance the capability of the original VoteNet to handle noise from background voting. The three proposed strategies all contribute to more effective voting and improved performance, resulting in a novel 3D object detector, termed VENet. Experiments show that our method outperforms state-of-the-art methods on benchmark datasets. Ablation studies demonstrate the effectiveness of the proposed components.

In this paper, we address the 3D object detection task by capturing multi-level contextual information with the self-attention mechanism and multi-scale feature fusion. Most existing 3D object detection methods recognize objects individually, without giving any consideration on contextual information between these objects. Comparatively, we propose Multi-Level Context VoteNet (MLCVNet) to recognize 3D objects correlatively, building on the state-of-the-art VoteNet. We introduce three context modules into the voting and classifying stages of VoteNet to encode contextual information at different levels. Specifically, a Patch-to-Patch Context (PPC) module is employed to capture contextual information between the point patches, before voting for their corresponding object centroid points. Subsequently, an Object-to-Object Context (OOC) module is incorporated before the proposal and classification stage, to capture the contextual information between object candidates. Finally, a Global Scene Context (GSC) module is designed to learn the global scene context. We demonstrate these by capturing contextual information at patch, object and scene levels. Our method is an effective way to promote detection accuracy, achieving new state-of-the-art detection performance on challenging 3D object detection datasets, i.e., SUN RGBD and ScanNet. We also release our code at https://github.com/NUAAXQ/MLCVNet.

Video and image sources are frequently applied in the area of defect inspection in industrial community. For the recognition and classification of sewer defects, a significant number of videos and images of sewers are collected. These data are then checked by human and some traditional methods to recognize and classify the sewer defects, which is inefficient and error-prone. Previously developed features like SIFT are unable to comprehensively represent such defects. Therefore, feature representation is especially important for defect autoclassification. In this paper, we study the automatic extraction of feature representation for sewer defects via deep learning. Moreover, a complete automatic system for classifying sewer defects is proposed built on a two-level hierarchical deep convolutional neural network, which shows high performance with respect to classification accuracy. The proposed network is trained on a novel data set with over 40 000 sewer images. The system has been successfully applied in the practical production, confirming its robustness and feasibility to real-world applications. The source code and trained model are available at the project website.1

RGB-D video segmentation is important for many applications, including scene understanding, object tracking, and robotic grasping. However, to segment RGB-D frames over a long video sequence into globally consistent segmentation is still a challenging problem. Current methods often lose pixel correspondences between frames under occlusion and, thus, fail to generate consistent and continuous segmentation results. To address this problem, we propose a novel spatiotemporal RGB-D video segmentation framework that automatically segments and tracks objects with continuity and consistency over time. Our approach first produces consistent segments in some keyframes by region clustering, and then propagates the segmentation result to a whole video sequence via a mask propagation scheme in bilateral space. Instead of exploiting local optical, flow information to establish correspondences between adjacent frames, we leverage scale-invariant feature transform (SIFT) flow and bilateral representation to solve inconsistency under occlusion. Moreover, our method automatically extracts multiple objects of interest and tracks them without any user input hint. A variety of experiments demonstrates effectiveness and robustness of our proposed method.

Abstract We present a sparse optimization framework for extracting sparse shape priors from a collection of 3D models. Shape priors are defined as point‐set neighborhoods sampled from shape surfaces which convey important information encompassing normals and local shape characterization. A 3D shape model can be considered to be formed with a set of 3D local shape priors, while most of them are likely to have similar geometry. Our key observation is that the local priors extracted from a family of 3D shapes lie in a very low‐dimensional manifold. Consequently, a compact and informative subset of priors can be learned to efficiently encode all shapes of the same family. A comprehensive library of local shape priors is first built with the given collection of 3D models of the same family. We then formulate a global, sparse optimization problem which enforces selecting representative priors while minimizing the reconstruction error. To solve the optimization problem, we design an efficient solver based on the Augmented Lagrangian Multipliers method (ALM). Extensive experiments exhibit the power of our data‐driven sparse priors in elegantly solving several high‐level shape analysis applications and geometry processing tasks, such as shape retrieval, style analysis and symmetry detection.