Dominant Set Clustering and Pooling for Multi-View 3D Object Recognition

Wang, Chu; Pelillo, Marcello; Siddiqi, Kaleem

doi:10.5244/c.31.64

Cited by 110 publications

(125 citation statements)

References 16 publications

(75 reference statements)

Supporting

Mentioning

124

Contrasting

Unclassified

Order By: Relevance

“…Shapebased methods cover 3DShapeNets [26], VoxNet [27], Sub-VolSup [13], AniProbing [13], VRN & VRN-Ensemble [28], FPNN [29], PointNet [30], PointNet++ [31], O-CNN [32] and So-Net [33]. The view-based methods include MVCNN [12], DeepPano [34], PANORAMA-NN [35], PANORAMA-ENN [36], MVCNN-Alex [13], MVCNN-MultiRes [13], Pairwise [16], DomSetClust [14], RotationNet [17] Table 2 gives the detail comparison against state-of-theart discrete view-based methods with their different processing strategies. Apparently, RotationNet, DomSetClust and MVCNN-MultiRes get the better performance.…”

Section: Comparison With State-of-the-artsmentioning

confidence: 99%

VERAM: View-Enhanced Recurrent Attention Model for 3D Shape Classification

Chen

Zheng

Zhang

et al. 2019

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

Multi-view deep neural network is perhaps the most successful approach in 3D shape classification. However, the fusion of multi-view features based on max or average pooling lacks a view selection mechanism, limiting its application in, e.g., multi-view active object recognition by a robot. This paper presents VERAM, a recurrent attention model capable of actively selecting a sequence of views for highly accurate 3D shape classification. VERAM addresses an important issue commonly found in existing attention-based models, i.e., the unbalanced training of the subnetworks corresponding to next view estimation and shape classification. The classification subnetwork is easily overfitted while the view estimation one is usually poorly trained, leading to a suboptimal classification performance. This is surmounted by three essential view-enhancement strategies: 1) enhancing the information flow of gradient backpropagation for the view estimation subnetwork, 2) devising a highly informative reward function for the reinforcement training of view estimation and 3) formulating a novel loss function that explicitly circumvents view duplication. Taking grayscale image as input and AlexNet as CNN architecture, VERAM with 9 views achieves instance-level and class-level accuracy of 95.5% and 95.3% on ModelNet10, 93.7% and 92.1% on ModelNet40, both are the state-of-the-art performance under the same number of views.

show abstract

Section: Comparison With State-of-the-artsmentioning

confidence: 99%

VERAM: View-Enhanced Recurrent Attention Model for 3D Shape Classification

Chen

Zheng

Zhang

et al. 2019

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

show abstract

“…High performance is achieved by taking advantage of the progress made in 2D classification by using a collection of 2D views of the object. They either pool over views [23], [24], apply more advanced schemes such as intelligently clustering before pooling [29] or jointly learning the corresponding object pose [13]. Beyond the power of such representations, they also take advantage of networks pretrained on large scale 2D datasets.…”

Section: A Geometric Deep Learningmentioning

confidence: 99%

Addressing the Sim2Real Gap in Robotic 3-D Object Classification

Weibel

Patten

Vincze

2020

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

Object classification with 3D data is an essential component of any scene understanding method. It has gained significant interest in a variety of communities, most notably in robotics and computer graphics. While the advent of deep learning has progressed the field of 3D object classification, most work using this data type are solely evaluated on CAD model datasets. Consequently, current work does not address the discrepancies existing between real and artificial data. In this work, we examine this gap in a robotic context by specifically addressing the problem of classification when transferring from artificial CAD models to real reconstructed objects. This is performed by training on ModelNet (CAD models) and evaluating on ScanNet (reconstructed objects). We show that standard methods do not perform well in this task. We thus introduce a method that carefully samples object parts that are reproducible under various transformations and hence robust. Using graph convolution to classify the composed graph of parts, our method significantly improves upon the baseline. -Code will be made publicly available on acceptance. -

show abstract

“…The penalty λ of CorrReg is set as 5e −4 . We report in Table XII results of MvFusion-Net without or with CorrReg regularization, where we also compare with recent state-of-the-art results [73], [60], [61] on ModelNet40 whose multi-view images are prepared following the same style of 1 st camera set-up in [59] (i.e., 12 camera views pointing towards upright orientation of object models). Due to varying architectural designs, network optimizers, and feature aggregation schemes, results of different methods in Table XII may not be directly comparable; nevertheless, it confirms the efficacy of CorrReg for better feature learning and aggregation from multiple views of 3D object shapes.…”

Section: Multi-view Recognition Of 3d Object Shapesmentioning

confidence: 99%

Deep Multi-View Learning Using Neuron-Wise Correlation-Maximizing Regularizers

Jia

Lin

Tan

et al. 2019

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Many machine learning problems concern with discovering or associating common patterns in data of multiple views or modalities. Multi-view learning is of the methods to achieve such goals. Recent methods propose deep multi-view networks via adaptation of generic Deep Neural Networks (DNNs), which concatenate features of individual views at intermediate network layers (i.e., fusion layers). In this work, we study the problem of multi-view learning in such end-to-end networks. We take a regularization approach via multi-view learning criteria, and propose a novel, effective, and efficient neuron-wise correlationmaximizing regularizer. We implement our proposed regularizers collectively as a correlation-regularized network layer (CorrReg). CorrReg can be applied to either fully-connected or convolutional fusion layers, simply by replacing them with their CorrReg counterparts. By partitioning neurons of a hidden layer in generic DNNs into multiple subsets, we also consider a multi-view feature learning perspective of generic DNNs. Such a perspective enables us to study deep multi-view learning in the context of regularized network training, for which we present control experiments of benchmark image classification to show the efficacy of our proposed CorrReg. To investigate how CorrReg is useful for practical multi-view learning problems, we conduct experiments of RGB-D object/scene recognition and multi-view based 3D object recognition, using networks with fusion layers that concatenate intermediate features of individual modalities or views for subsequent classification. Applying CorrReg to fusion layers of these networks consistently improves classification performance. In particular, we achieve the new state of the art on the benchmark RGB-D object and RGB-D scene datasets. We make the implementation of CorrReg publicly available.

show abstract

Dominant Set Clustering and Pooling for Multi-View 3D Object Recognition

Cited by 110 publications

References 16 publications

VERAM: View-Enhanced Recurrent Attention Model for 3D Shape Classification

VERAM: View-Enhanced Recurrent Attention Model for 3D Shape Classification

Addressing the Sim2Real Gap in Robotic 3-D Object Classification

Deep Multi-View Learning Using Neuron-Wise Correlation-Maximizing Regularizers

Contact Info

Product

Resources

About