Deep Learning-Based Point Cloud Geometry Coding: RD Control Through Implicit and Explicit Quantization

Guarda, Andre F. R.; Rodrigues, Nuno M. M.; Pereira, Fernando

doi:10.1109/icmew46912.2020.9106022

Cited by 26 publications

(18 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other approaches were presented by Guarda et al in [ 108 ]. Initially, the authors proposed in [ 109 ] a simple convolutional network where the latent space is quantized in order to perform entropy coding.…”

Section: Compressionmentioning

confidence: 99%

“…Initially, the authors proposed in [ 109 ] a simple convolutional network where the latent space is quantized in order to perform entropy coding. Later in [ 108 ], the approach was improved by implementing the hyperpriors technique [ 95 ] and by proposing an implicit/explicit quantization framework (see Figure 18 ). In implicit quantization, a deep learning model is optimized to minimize a given rate–distortion tradeoff while, in the explicit one, the latent space is quantized with different step values depending on the required quality/rate tradeoff.…”

Section: Compressionmentioning

confidence: 99%

See 1 more Smart Citation

Recent Advancements in Learning Algorithms for Point Clouds: An Updated Overview

Camuffo

Mari

Milani

2022

Sensors

View full text Add to dashboard Cite

Recent advancements in self-driving cars, robotics, and remote sensing have widened the range of applications for 3D Point Cloud (PC) data. This data format poses several new issues concerning noise levels, sparsity, and required storage space; as a result, many recent works address PC problems using Deep Learning (DL) solutions thanks to their capability to automatically extract features and achieve high performances. Such evolution has also changed the structure of processing chains and posed new problems to both academic and industrial researchers. The aim of this paper is to provide a comprehensive overview of the latest state-of-the-art DL approaches for the most crucial PC processing operations, i.e., semantic scene understanding, compression, and completion. With respect to the existing reviews, the work proposes a new taxonomical classification of the approaches, taking into account the characteristics of the acquisition set up, the peculiarities of the acquired PC data, the presence of side information (depending on the adopted dataset), the data formatting, and the characteristics of the DL architectures. This organization allows one to better comprehend some final performance comparisons on common test sets and cast a light on the future research trends.

show abstract

Section: Compressionmentioning

confidence: 99%

Section: Compressionmentioning

confidence: 99%

Recent Advancements in Learning Algorithms for Point Clouds: An Updated Overview

Camuffo

Mari

Milani

2022

Sensors

View full text Add to dashboard Cite

show abstract

“…Most of the learning-based models search for an optimum nonlinear transformation in terms of rate distortion trade-off during training time by using an appropriate differentiable objective function. For instance, the estimated entropy of a quantized latent vector can be used along with the reconstruction loss in an objective function [32,33]. Within the image domain, convolutional neural networks (CNNs) are used to extract the translationinvariant features that encapsulate spatial correlations among neighboring pixels.…”

Section: Learning-based Compressionmentioning

confidence: 99%

Variable Rate Compression for Raw 3D Point Clouds

Al¹,

Beksi²

2022

Preprint

View full text Add to dashboard Cite

In this paper, we propose a novel variable rate deep compression architecture that operates on raw 3D point cloud data. The majority of learning-based point cloud compression methods work on a downsampled representation of the data. Moreover, many existing techniques require training multiple networks for different compression rates to generate consolidated point clouds of varying quality. In contrast, our network is capable of explicitly processing point clouds and generating a compressed description at a comprehensive range of bitrates. Furthermore, our approach ensures that there is no loss of information as a result of the voxelization process and the density of the point cloud does not affect the encoder/decoder performance. An extensive experimental evaluation shows that our model obtains state-of-the-art results, it is computationally efficient, and it can work directly with point cloud data thus avoiding an expensive voxelized representation.

show abstract

“…The impact of several parameters added to the initial version of the former network is evaluated in [13]. In [14], a rate-distortion performance analysis is conducted on the latent space of [12], which is enriched with a hyper-prior and the possibility of explicit quantization in [15]. In [16], a deeper auto-encoding architecture is proposed, based on 3D convolution layers stacked with Voxception-ResNet structures and a hyper-prior.…”

Section: Related Workmentioning

confidence: 99%

On Block Prediction For Learning-Based Point Cloud Compression

Lazzarotto

Alexiou

Ebrahimi

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Point clouds are among popular visual representations for immersive media. However, the vast amount of information generated during their acquisition requires effective compression for practical applications. Although relevant activities from standardization bodies have led to state-of-the-art compression using conventional methods, learning-based encoders have recently emerged as promising solutions with comparable performance while offering additional attractive features. Yet, there is still a large unexplored space for research that can lead to further advances. In this paper, we propose a block prediction module for bit-rate reduction of geometryonly point clouds. Our method exploits spatial redundancies at the decoding stage between block partitions in the point cloud, and predicts a query block using Generative Adversarial Networks. Results show performance improvements of the objective metrics at low bit-rates, after integration in a baseline auto-encoder architecture.

show abstract

Deep Learning-Based Point Cloud Geometry Coding: RD Control Through Implicit and Explicit Quantization

Cited by 26 publications

References 13 publications

Recent Advancements in Learning Algorithms for Point Clouds: An Updated Overview

Recent Advancements in Learning Algorithms for Point Clouds: An Updated Overview

Variable Rate Compression for Raw 3D Point Clouds

On Block Prediction For Learning-Based Point Cloud Compression

Contact Info

Product

Resources

About