Handling Data Scarcity Through Data Augmentation in Training of Deep Neural Networks for 3D Data Processing

Abstract

Due to the availability of cheap 3D sensors such as Kinect and LiDAR, the use of 3D data in various domains such as manufacturing, healthcare, and retail to achieve operational safety, improved outcomes, and enhanced customer experience has gained momentum in recent years. In many of these domains, object recognition is being performed using 3D data against the difficulties posed by illumination, pose variation, scaling, etc present in 2D data. In this work, we propose three data augmentation techniques for 3D data in point cloud representation that use sub-sampling. We then verify that the 3D samples created through data augmentation carry the same information by comparing the Iterative Closest Point Registration Error within the sub-samples, between the sub-samples and their parent sample, between the sub-samples with different parents and the same subject, and finally, between the sub-samples of different subjects. We also verify that the augmented sub-samples have the same characteristics and features as those of the original 3D point cloud by applying the Central Limit Theorem.

Car Emotion Labeling Based on Color-SSL Semi-Supervised Learning Algorithm by Color Augmentation

“…Sedik et al used the image rotation method to augment the data of X-ray chest images, achieving excellent performance and efectively solving the problem of deep learning in detecting and screening COVID-19 [44]. Srivastava et al used three sampling methods to augment 3D data to solve the scarcity of 3D data in depth neural networks [45]. Rahman et al used GAN to augment the COVID-19 dataset and verifed the accuracy of the test with the deep learning model [46].…”

A deep learning model for online doctor rating prediction

“…W T and W C are the corresponding weight matrices of the transform and carry gate, respectively. To allow the carry function C to hold unmodified information, the bias b T (a hyperparameter tuned through grid search) of transform gate is initialized as a negative number (Srivastava et al, 2015). The other hyperparameter is the number of layers in the highway network, which is also optimized through grid search.…”

“…The research question therefore is: How to effectively predict the doctor ratings from online textual reviews? We do so by proposing a novel deep-learning model, namely, ODRP-HABiLSTM, consisting of (a) selfattention mechanism in combination with (b) bidirectional long-short term memory network (BiLSTM) and (c) highway network (Srivastava et al, 2015). The model constructs two separate F I G U R E 1 Data sparsity and the need for rating prediction.…”

A novel framework for detecting non‐recurrent road traffic anomalies by combining a temporal graph convolutional network and hierarchical time memory detector

“…With the improvement of the computing power of the hardware, machine learning models, especially DL models have demonstrated strong spatiotemporal feature mining abilities and have been widely used in various applications of geographic information science (GIS), such as hospital site selection studies (Benmoussa et al, 2022), generation of crowd counting networks (Hao et al, 2022), and processing of 3D data (Srivastava et al, 2022). Many machine learning models have been applied to mine the spatiotemporal features for traffic flow prediction (Fusco et al, 2016; Yi et al, 2017), and an increasing number of DL models have been used to predict traffic parameters (Cui et al, 2018; Polson & Sokolov, 2017).…”