2020
DOI: 10.1007/978-3-030-58523-5_16
|View full text |Cite
|
Sign up to set email alerts
|

An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
43
0

Year Published

2021
2021
2022
2022

Publication Types

Select...
5
3
1

Relationship

0
9

Authors

Journals

citations
Cited by 77 publications
(43 citation statements)
references
References 42 publications
0
43
0
Order By: Relevance
“…This increases the ability of such layers to accurately model spatial and temporal interactions among pedestrians. Conversely, in [55] the authors adapt an arrangement of LSTM layers to process sparse 3D data structures as opposed to changing internal data representation. The authors choose to model the interactions among pedestrians using graphs and, consequently, graph convolutional networks.…”
Section: Lstm-based Methodsmentioning
confidence: 99%
See 1 more Smart Citation
“…This increases the ability of such layers to accurately model spatial and temporal interactions among pedestrians. Conversely, in [55] the authors adapt an arrangement of LSTM layers to process sparse 3D data structures as opposed to changing internal data representation. The authors choose to model the interactions among pedestrians using graphs and, consequently, graph convolutional networks.…”
Section: Lstm-based Methodsmentioning
confidence: 99%
“…-cannot deduce interactions without additional mechanisms -generally more difficult to train Models based on LSTM cell configurations that directly predict vehicle/obstacle occupancy [44,49] Model for motion prediction based on vehicle maneuvers [45] LSTM-based architectures that generate appearance and motion models and learn interaction information over extended sequences [43,47,51] Multi-layer GRU-based architecture which splits and reconnects tracklets generated from convolutional features [46] Basic RNN that encodes information from multiple frame sequences [42,48] LSTM layers that focus on learning and interpreting actor intentions [52,53] LSTM-based object detection and tracking adapted for sequences of higher-dimensional data [55] LSTM models that encode relationships between actors using graph representations [56,57] LSTM model that uses multidimensional internal representations of data sequences [54] Methods not relying on neural networks…”
Section: Strengthsmentioning
confidence: 99%
“…The literature [28] conducted a study on the robustness of point cloud 3D target detection. A sparse LSTM multi-frame 3D target detection algorithm focusing on the time-domain information of point clouds was proposed in the literature [29]. All these studies provide new research perspectives for point cloud target detection.…”
Section: Related Workmentioning
confidence: 99%
“…Multimodal sensor data combination improves the description capacity by generative adversarial networks (GAN), and variation auto-encoder (VAE) [25,26]. The recurrent neural networks (RNN) with convolutional and LSTM recurrent units are utilized to measure and model the driving scene based on multi-model data [27,28], and DCNN has been considered to automate the feature extraction process [29,30]. MDF is used to guarantee the robustness of information assimilated from numerous sensors in various driving situations.…”
Section: Multimodal Data Fusion (Mdf)mentioning
confidence: 99%