A Computational Taxonomy of Deep Neural Networks for 3D Point Cloud Processing

Home

Discover

Bookmarks

Keep Up With Latest Trending Papers. Computer Science, AI and Machine Learning and more.Subscribe

Neural and Evolutionary Computing

Robotics

A Computational Taxonomy of Deep Neural Networks for 3D Point Cloud Processing

Analyzing Deep Learning Representations of Point Clouds for Real-Time In-Vehicle LiDAR Perception

Deep neural networks operating on 3-dimensional (3d) point clouds of data from multiple high-resolution lidar sensors must extract semantics from this increasingly precise picture of the vehicle s environment.

However, it is computationally difficult to make use of the ever-increasing amounts of data from multiple high-resolution lidar sensors.

Authors

Marc Uecker, Tobias Fleck, Marcel Pflugfelder, J. Marius Zöllner

Related Topics

Deep neural networks

Deep learning approaches

Neural networks

Artificial intelligence

Deep learning methods

Data representation

Point cloud representations

Computational representations

Taxonomy of lidar point cloud representations

Neural point cloud processing methods

Method of feature aggregation

Kissme

Lidar point cloud representation

Introduction

In this work, we present a taxonomy of different architecture designs, based on design decisions regarding the point cloud data representation.

This taxonomy is used to analyze the impact of these design decisions on the run-time performance characteristics in.

Based on this analysis, we also provide insights and recommendations for future work in.

Result

In this work, we present a taxonomy covering the most significant design decisions regarding the lidar point cloud representation used in deep neural networks.

By categorizing recent state-of-the-art methods for 3d semantic segmentation, we uncover common trends which can be traced back to common design decisions regarding representational capacity.

We dive deeply into the design decisions made during the development of point cloud representations, and analyze the impact of each design decision on memory consumption, run-time latency and representational capacity during the design phase of the model.

Finally, we discuss for each of the limitations present in the autonomous vehicle use-case, which design decisions can be altered to lighten the load of latency-or memory-constraints, or increase representational capacity.

Read the Paper

Read Paper

Bookmarks

Bookmarks

Keep Up With Latest Trending Papers. Computer Science, AI and Machine Learning and more.Subscribe

Neural and Evolutionary Computing

Robotics

A Computational Taxonomy of Deep Neural Networks for 3D Point Cloud Processing

Analyzing Deep Learning Representations of Point Clouds for Real-Time In-Vehicle LiDAR Perception

However, it is computationally difficult to make use of the ever-increasing amounts of data from multiple high-resolution lidar sensors.

Authors

Marc Uecker, Tobias Fleck, Marcel Pflugfelder, J. Marius Zöllner

Related Topics

Deep neural networks

Deep learning approaches

Neural networks

Artificial intelligence

Deep learning methods

Data representation

Point cloud representations

Computational representations

Taxonomy of lidar point cloud representations

Neural point cloud processing methods

Method of feature aggregation

Kissme

Lidar point cloud representation

Introduction

In this work, we present a taxonomy of different architecture designs, based on design decisions regarding the point cloud data representation.

This taxonomy is used to analyze the impact of these design decisions on the run-time performance characteristics in.

Based on this analysis, we also provide insights and recommendations for future work in.

Result

In this work, we present a taxonomy covering the most significant design decisions regarding the lidar point cloud representation used in deep neural networks.

By categorizing recent state-of-the-art methods for 3d semantic segmentation, we uncover common trends which can be traced back to common design decisions regarding representational capacity.

Read the Paper

Read Paper

Home

Discover

Bookmarks

◐ Latest Activity

◐ Recommended Follows