In this paper, we propose Attention Based Decomposition Network (ABD-Net), for point cloud decomposition into basic geometric shapes namely, plane, sphere, cone and cylinder. We show improved performance of 3D object classification using attention features based on primitive shapes in point clouds. Point clouds, being the simple and compact representation of 3D objects have gained increasing popularity. They demand robust methods for feature extraction due to unorderness in point sets. In ABD-Net the proposed Local Proximity Encapsulator captures the local geometric variations along with spatial encoding around each point from the input point sets. The encapsulated local features are further passed to proposed Attention Feature Encoder to learn basic shapes in point cloud. Attention Feature Encoder models geometric relationship between the neighborhoods of all the points resulting in capturing global point cloud information. We demonstrate the results of our proposed ABD-Net on ANSI mechanical component and ModelNet40 datasets. We also demonstrate the effectiveness of ABD-Net over the acquired attention features by improving the performance of 3D object classification on ModelNet40 benchmark dataset and compare them with state-of-the-art techniques.
3D point cloud analysis has drawn a lot of research attention due to its wide applications. However, collecting massive labelled 3D point cloud data is both time-consuming and labor-intensive. This calls for data-efficient learning methods. In this work we propose PointDisc, a point discriminative learning method to leverage self-supervisions for data-efficient 3D point cloud classification and segmentation. PointDisc imposes a novel point discrimination loss on the middle and global level features produced by the backbone network. This point discrimination loss enforces learned features to be consistent with points belonging to the corresponding local shape region and inconsistent with randomly sampled noisy points. We conduct extensive experiments on 3D object classification, 3D semantic and part segmentation, showing the benefits of PointDisc for data-efficient learning. Detailed analysis demonstrate that PointDisc learns unsupervised features that well capture local and global geometry.
This paper presents new designs of graph convolutional neural networks (GCNs) on 3D meshes for 3D object segmentation and classification. We use the faces of the mesh as basic processing units and represent a 3D mesh as a graph where each node corresponds to a face. To enhance the descriptive power of the graph, we introduce a 1-ring face neighbourhood structure to derive novel multi-dimensional spatial and structure features to represent the graph nodes. Based on this new graph representation, we then design a densely connected graph convolutional block which aggregates local and regional features as the key construction component to build effective and efficient practical GCN models for 3D object classification and segmentation. We will present experimental results to show that our new technique outperforms state of the art where our models are shown to have the smallest number of parameters and consietently achieve the highest accuracies across a number of benchmark datasets. We will also present ablation studies to demonstrate the soundness of our design principles and the effectiveness of our practical models.
Self-attention mechanism recently achieves impressive advancement in Natural Language Processing (NLP) and Image Processing domains. And its permutation invariance property makes it ideally suitable for point cloud processing. Inspired by this remarkable success, we propose an end-to-end architecture, dubbed Cross-Level Cross-Scale Cross-Attention Network (CLCSCANet), for point cloud representation learning. First, a point-wise feature pyramid module is introduced to hierarchically extract features from different scales or resolutions. Then a cross-level cross-attention is designed to model long-range inter-level and intra-level dependencies. Finally, we develop a cross-scale cross-attention module to capture interactions between-and-within scales for representation enhancement. Compared with state-of-the-art approaches, our network can obtain competitive performance on challenging 3D object classification, point cloud segmentation tasks via comprehensive experimental evaluation.
Scene graphs are a compact and explicit representation successfully used in a variety of 2D scene understanding tasks. This work proposes a method to incrementally build up semantic scene graphs from a 3D environment given a sequence of RGB-D frames. To this end, we aggregate PointNet features from primitive scene components by means of a graph neural network. We also propose a novel attention mechanism well suited for partial and missing graph data present in such an incremental reconstruction scenario. Although our proposed method is designed to run on submaps of the scene, we show it also transfers to entire 3D scenes. Experiments show that our approach outperforms 3D scene graph prediction methods by a large margin and its accuracy is on par with other 3D semantic and panoptic segmentation methods while running at 35 Hz.
While object semantic understanding is essential for most service robotic tasks, 3D object classification is still an open problem. Learning from artificial 3D models alleviates the cost of annotation necessary to approach this problem, but most methods still struggle with the differences existing between artificial and real 3D data. We conjecture that the cause of those issue is the fact that many methods learn directly from point coordinates, instead of the shape, as the former is hard to center and to scale under variable occlusions reliably. We introduce spherical kernel point convolutions that directly exploit the object surface, represented as a graph, and a voting scheme to limit the impact of poor segmentation on the classification results. Our proposed approach improves upon state-of-the-art methods by up to 36% when transferring from artificial objects to real objects.
3D object representation learning is a fundamental challenge in computer vision to infer about the 3D world. Recent advances in deep learning have shown their efficiency in 3D object recognition, among which view-based methods have performed best so far. However, feature learning of multiple views in existing methods is mostly performed in a supervised fashion, which often requires a large amount of data labels with high costs. In contrast, self-supervised learning aims to learn multi-view feature representations without involving labeled data. To this end, we propose a novel self-supervised paradigm to learn Multi-View Transformation Equivariant Representations (MV-TER), exploring the equivariant transformations of a 3D object and its projected multiple views. Specifically, we perform a 3D transformation on a 3D object, and obtain multiple views before and after the transformation via projection. Then, we self-train a representation to capture the intrinsic 3D object representation by decoding 3D transformation parameters from the fused feature representations of multiple views before and after the transformation. Experimental results demonstrate that the proposed MV-TER significantly outperforms the state-of-the-art view-based approaches in 3D object classification and retrieval tasks, and show the generalization to real-world datasets.
Data augmentation is an effective regularization strategy to alleviate the overfitting, which is an inherent drawback of the deep neural networks. However, data augmentation is rarely considered for point cloud processing despite many studies proposing various augmentation methods for image data. Actually, regularization is essential for point clouds since lack of generality is more likely to occur in point cloud due to small datasets. This paper proposes a Rigid Subset Mix (RSMix), a novel data augmentation method for point clouds that generates a virtual mixed sample by replacing part of the sample with shape-preserved subsets from another sample. RSMix preserves structural information of the point cloud sample by extracting subsets from each sample without deformation using a neighboring function. The neighboring function was carefully designed considering unique properties of point cloud, unordered structure and non-grid. Experiments verified that RSMix successfully regularized the deep neural networks with remarkable improvement for shape classification. We also analyzed various combinations of data augmentations including RSMix with single and multi-view evaluations, based on abundant ablation studies.
Convolutional neural networks (CNNs) have been widely used in various vision tasks, e.g. image classification, semantic segmentation, etc. Unfortunately, standard 2D CNNs are not well suited for spherical signals such as panorama images or spherical projections, as the sphere is an unstructured grid. In this paper, we present Spherical Transformer which can transform spherical signals into vectors that can be directly processed by standard CNNs such that many well-designed CNNs architectures can be reused across tasks and datasets by pretraining. To this end, the proposed method first uses local structured sampling methods such as HEALPix to construct a transformer grid by using the information of spherical points and its adjacent points, and then transforms the spherical signals to the vectors through the grid. By building the Spherical Transformer module, we can use multiple CNN architectures directly. We evaluate our approach on the tasks of spherical MNIST recognition, 3D object classification and omnidirectional image semantic segmentation. For 3D object classification, we further propose a rendering-based projection method to improve the performance and a rotational-equivariant model to improve the anti-rotation ability. Experimental results on three tasks show that our approach achieves superior performance over state-of-the-art methods.
3D data that contains rich geometry information of objects and scenes is valuable for understanding 3D physical world. With the recent emergence of large-scale 3D datasets, it becomes increasingly crucial to have a powerful 3D generative model for 3D shape synthesis and analysis. This paper proposes a deep 3D energy-based model to represent volumetric shapes. The maximum likelihood training of the model follows an "analysis by synthesis" scheme. The benefits of the proposed model are six-fold: first, unlike GANs and VAEs, the model training does not rely on any auxiliary models; second, the model can synthesize realistic 3D shapes by Markov chain Monte Carlo (MCMC); third, the conditional model can be applied to 3D object recovery and super resolution; fourth, the model can serve as a building block in a multi-grid modeling and sampling framework for high resolution 3D shape synthesis; fifth, the model can be used to train a 3D generator via MCMC teaching; sixth, the unsupervisedly trained model provides a powerful feature extractor for 3D data, which is useful for 3D object classification. Experiments demonstrate that the proposed model can generate high-quality 3D shape patterns and can be useful for a wide variety of 3D shape analysis.
In 2D image processing, some attempts decompose images into high and low frequency components for describing edge and smooth parts respectively. Similarly, the contour and flat area of 3D objects, such as the boundary and seat area of a chair, describe different but also complementary geometries. However, such investigation is lost in previous deep networks that understand point clouds by directly treating all points or local patches equally. To solve this problem, we propose Geometry-Disentangled Attention Network (GDANet). GDANet introduces Geometry-Disentangle Module to dynamically disentangle point clouds into the contour and flat part of 3D objects, respectively denoted by sharp and gentle variation components. Then GDANet exploits Sharp-Gentle Complementary Attention Module that regards the features from sharp and gentle variation components as two holistic representations, and pays different attentions to them while fusing them respectively with original point cloud features. In this way, our method captures and refines the holistic and complementary 3D geometric semantics from two distinct disentangled components to supplement the local information. Extensive experiments on 3D object classification and segmentation benchmarks demonstrate that GDANet achieves the state-of-the-arts with fewer parameters. Code is released on https://github.com/mutianxu/GDANet.
Object point cloud classification has drawn great research attention since the release of benchmarking datasets, such as the ModelNet and the ShapeNet. These benchmarks assume point clouds covering complete surfaces of object instances, for which plenty of high-performing methods have been developed. However, their settings deviate from those often met in practice, where, due to (self-)occlusion, a point cloud covering partial surface of an object is captured from an arbitrary view. We show in this paper that performance of existing point cloud classifiers drops drastically under the considered single-view, partial setting; the phenomenon is consistent with the observation that semantic category of a partial object surface is less ambiguous only when its distribution on the whole surface is clearly specified. To this end, we argue for a single-view, partial setting where supervised learning of object pose estimation should be accompanied with classification. Technically, we propose a baseline method of Pose-Accompanied Point cloud classification Network (PAPNet); built upon SE(3)-equivariant convolutions, the PAPNet learns intermediate pose transformations for equivariant features defined on vector fields, which makes the subsequent classification easier (ideally) in the category-level, canonical pose. By adapting existing ModelNet40 and ScanNet datasets to the single-view, partial setting, experiment results can verify the necessity of object pose estimation and superiority of our PAPNet to existing classifiers.
3D object classification has attracted appealing attentions in academic researches and industrial applications. However, most existing methods need to access the training data of past 3D object classes when facing the common real-world scenario: new classes of 3D objects arrive in a sequence. Moreover, the performance of advanced approaches degrades dramatically for past learned classes (i.e., catastrophic forgetting), due to the irregular and redundant geometric structures of 3D point cloud data. To address these challenges, we propose a new Incremental 3D Object Learning (i.e., I3DOL) model, which is the first exploration to learn new classes of 3D object continually. Specifically, an adaptive-geometric centroid module is designed to construct discriminative local geometric structures, which can better characterize the irregular point cloud representation for 3D object. Afterwards, to prevent the catastrophic forgetting brought by redundant geometric information, a geometric-aware attention mechanism is developed to quantify the contributions of local geometric structures, and explore unique 3D geometric characteristics with high contributions for classes incremental learning. Meanwhile, a score fairness compensation strategy is proposed to further alleviate the catastrophic forgetting caused by unbalanced data between past and new classes of 3D object, by compensating biased prediction for new classes in the validation phase. Experiments on 3D representative datasets validate the superiority of our I3DOL framework.
Voxel-based 3D object classification has been thoroughly studied in recent years. Most previous methods convert the classic 2D convolution into a 3D form that will be further applied to objects with binary voxel representation for classification. However, the binary voxel representation is not very effective for 3D convolution in many cases. In this paper, we propose a hybrid cascade architecture for voxel-based 3D object classification. It consists of three stages composed of fully connected and convolutional layers, dealing with easy, moderate, and hard 3D models respectively. Both accuracy and speed can be balanced in our proposed method. By giving each voxel a signed distance value, an obvious gain regarding the accuracy can be observed. Besides, the mean inference time can be speeded up hugely compared with the state-of-the-art point cloud and voxel based methods.
In this work, we present Point Transformer, a deep neural network that operates directly on unordered and unstructured point sets. We design Point Transformer to extract local and global features and relate both representations by introducing the local-global attention mechanism, which aims to capture spatial point relations and shape information. For that purpose, we propose SortNet, as part of the Point Transformer, which induces input permutation invariance by selecting points based on a learned score. The output of Point Transformer is a sorted and permutation invariant feature list that can directly be incorporated into common computer vision applications. We evaluate our approach on standard classification and part segmentation benchmarks to demonstrate competitive results compared to the prior work. Code is publicly available at: https://github.com/engelnico/point-transformer
Point clouds are often sparse and incomplete, which imposes difficulties for real-world applications. Existing shape completion methods tend to generate rough shapes without fine-grained details. Considering this, we introduce a two-branch network for shape completion. The first branch is a cascaded shape completion sub-network to synthesize complete objects, where we propose to use the partial input together with the coarse output to preserve the object details during the dense point reconstruction. The second branch is an auto-encoder to reconstruct the original partial input. The two branches share a same feature extractor to learn an accurate global feature for shape completion. Furthermore, we propose two strategies to enable the training of our network when ground truth data are not available. This is to mitigate the dependence of existing approaches on large amounts of ground truth training data that are often difficult to obtain in real-world applications. Additionally, our proposed strategies are also able to improve the reconstruction quality for fully supervised learning. We verify our approach in self-supervised, semi-supervised and fully supervised settings with superior performances. Quantitative and qualitative results on different datasets demonstrate that our method achieves more realistic outputs than state-of-the-art approaches on the point cloud completion task.
Nonnegative matrix factorization (NMF) has become a ubiquitous tool for data analysis. An important variant is the sparse NMF problem which arises when we explicitly require the learnt features to be sparse. A natural measure of sparsity is the L$_0$ norm, however its optimization is NP-hard. Mixed norms, such as L$_1$/L$_2$ measure, have been shown to model sparsity robustly, based on intuitive attributes that such measures need to satisfy. This is in contrast to computationally cheaper alternatives such as the plain L$_1$ norm. However, present algorithms designed for optimizing the mixed norm L$_1$/L$_2$ are slow and other formulations for sparse NMF have been proposed such as those based on L$_1$ and L$_0$ norms. Our proposed algorithm allows us to solve the mixed norm sparsity constraints while not sacrificing computation time. We present experimental evidence on real-world datasets that shows our new algorithm performs an order of magnitude faster compared to the current state-of-the-art solvers optimizing the mixed norm and is suitable for large-scale datasets.
A key objective in multi-view learning is to model the information common to multiple parallel views of a class of objects/events to improve downstream learning tasks. In this context, two open research questions remain: How can we model hundreds of views per event? Can we learn robust multi-view embeddings without any knowledge of how these views are acquired? We present a neural method based on multi-view correlation to capture the information shared across a large number of views by subsampling them in a view-agnostic manner during training. To provide an upper bound on the number of views to subsample for a given embedding dimension, we analyze the error of the bootstrapped multi-view correlation objective using matrix concentration theory. Our experiments on spoken word recognition, 3D object classification and pose-invariant face recognition demonstrate the robustness of view bootstrapping to model a large number of views. Results underscore the applicability of our method for a view-agnostic learning setting.
Local and global patterns of an object are closely related. Although each part of an object is incomplete, the underlying attributes about the object are shared among all parts, which makes reasoning the whole object from a single part possible. We hypothesize that a powerful representation of a 3D object should model the attributes that are shared between parts and the whole object, and distinguishable from other objects. Based on this hypothesis, we propose to learn point cloud representation by bidirectional reasoning between the local structures at different abstraction hierarchies and the global shape without human supervision. Experimental results on various benchmark datasets demonstrate the unsupervisedly learned representation is even better than supervised representation in discriminative power, generalization ability, and robustness. We show that unsupervisedly trained point cloud models can outperform their supervised counterparts on downstream classification tasks. Most notably, by simply increasing the channel width of an SSG PointNet++, our unsupervised model surpasses the state-of-the-art supervised methods on both synthetic and real-world 3D object classification datasets. We expect our observations to offer a new perspective on learning better representation from data structures instead of human annotations for point cloud understanding.
We introduce FPConv, a novel surface-style convolution operator designed for 3D point cloud analysis. Unlike previous methods, FPConv doesn't require transforming to intermediate representation like 3D grid or graph and directly works on surface geometry of point cloud. To be more specific, for each point, FPConv performs a local flattening by automatically learning a weight map to softly project surrounding points onto a 2D grid. Regular 2D convolution can thus be applied for efficient feature learning. FPConv can be easily integrated into various network architectures for tasks like 3D object classification and 3D scene segmentation, and achieve comparable performance with existing volumetric-type convolutions. More importantly, our experiments also show that FPConv can be a complementary of volumetric convolutions and jointly training them can further boost overall performance into state-of-the-art results.