Object Classification from 3D Volumetric Data with 3D Capsule Networks

Kakillioglu, Burak; Ahmad, Ayesha; Velipasalar, Senem

doi:10.1109/globalsip.2018.8646333

Cited by 5 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The work proposed in this paper is different from, and improved compared to our previous work [10], [11] in multiple ways including (i) the development of a weight pruning approach based on ADMM and applying this approach to the proposed 3D CapsNet to significantly reduce the number of weights and the memory requirements of the model for embedded deployment, and increase the accuracy at the same time; (ii) introducing an optimization to the dynamic routing mechanism for faster computation while maintaining the classification accuracy; (iii) performing a comprehensive set of experiments comparing accuracy, number of weights and compression ratio on different datasets and showing the significant decrease in the number of weights together with the effect of weight pruning on the classification accuracy; (iv) comparison with additional existing work and a base model; (v) providing a detailed analysis by using different data splits showing the performance on decreasing amounts of training data.…”

Section: The Contributions Of This Paper Include the Followingmentioning

confidence: 74%

“…In our experiments, we use the 3D CapsNet architecture shown in Fig. 2, and improve the performance of our previous work [10] by incorporating ADMM-based weight pruning optimization and routing mechanism optimization, which are detailed in Sections IV and V-F, respectively. We also perform additional comparative experiments with the state-of-the art models and a base model on multiple datasets with different training/testing splits to show the significant decrease in the number of weights together with the effect of weight pruning on the classification accuracy.…”

Section: D Capsule Networkmentioning

confidence: 99%

“…In our preliminary work [10], [11], we introduced a 3D Capsule Network, referred to as 3D CapsNet, to perform object classification from 3D volumetric data. Despite its ability to extract features from 3D objects, the large model size of the 3D CapsNet hurdles its deployment on resourceconstrained platforms, such as mobile phones, IoT devices, and unmanned vehicles etc.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

3D Capsule Networks for Object Classification With Weight Pruning

et al. 2020

Self Cite

View full text Add to dashboard Cite

The proliferation of 3D sensors, due to the increased demand for 3D data, induced the 3D computer vision research in the last decade, and 3D data processing has gained a lot of interest. As in many other applications in computer vision, deep learning-based methods were quickly applied to 3D data classification and have become the state-of-the-art in this area. More recently, capsule networks, which are novel neural structures, have been introduced to enhance the ability of neural networks to better capture the parts-relationship, which yields more accurate classification with less training data. Moreover, deploying deep machine learning models on mobile platforms requires the models to be optimized due to limited memory and computational constraints. In this work, we propose methods to boost the accuracies of a standard 3D CNN-based and a Capsule Network-based classifier, help the training to better generalize the data distribution with limited data, and optimize the models for resource-constrained environments, such as mobile platforms. We also show that the introduction of capsules to 3D object classification pipeline improves the classification performance with limited training data, while a specifically optimized weight pruning method keeps the model compact enough for mobile deployment. Our broad spectrum of experiments show that proposed methods improve the performance of the base model while significantly reducing the memory and computation requirements.

show abstract

Section: The Contributions Of This Paper Include the Followingmentioning

confidence: 74%

Section: D Capsule Networkmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

3D Capsule Networks for Object Classification With Weight Pruning

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

A novel 3D shape classification algorithm: point-to-vector capsule network

Cao

2021

Neural Comput & Applic

View full text Add to dashboard Cite

CapsNets algorithm

Zhang

Wei

Han

2020

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Capsule networks (CapsNets) is a commonly used neural network structure, especially in computer vision research., which may have a significant impact on deep learning. Convolutional neural networks (CNNs) have reached superhuman level in various computer classification tasks such as classification, parsing, object detection, semantic segmentation, and instance segmentation. Traditional deep learning tends to deteriorate when the relevance of detection objects changes. For this kind of situation, this paper improves a new network, namely the capsule network(CapsNets), the network can achieve a higher detection rate in the hostile environment.

show abstract

Object Classification from 3D Volumetric Data with 3D Capsule Networks

Cited by 5 publications

References 12 publications

3D Capsule Networks for Object Classification With Weight Pruning

3D Capsule Networks for Object Classification With Weight Pruning

A novel 3D shape classification algorithm: point-to-vector capsule network

CapsNets algorithm

Contact Info

Product

Resources

About