Temporal LiDAR Frame Prediction for Autonomous Driving
作者： Avideh Zakhor , David Deng
DOI: 10.1109/3DV50981.2020.00093
关键词:
摘要: Anticipating the future in a dynamic scene is critical for many fields such as autonomous driving and robotics. In this paper we propose class of novel neural network architectures to predict LiDAR frames given previous ones. Since ground truth application simply next frame sequence, can train our models an self-supervised fashion. Our proposed are based on FlowNet3D Dynamic Graph CNN. We use Chamfer Distance (CD) Earth Mover’s (EMD) loss functions evaluation metrics. evaluate using newly released nuScenes dataset, characterize their performance complexity with several baselines. Compared directly FlowNet3D, achieve CD EMD nearly order magnitude lower. addition, show that predictions generate reasonable flow approximations without any labelled supervision.
arxiv.org
sci-hub.st 参考文章(22)
Dimitri Bertsekas, A distributed asynchronous relaxation algorithm for the assignment problem conference on decision and control. ,vol. 24, pp. 1703- 1704 ,(1985) , 10.1109/CDC.1985.268826
Daniel Maturana, Sebastian Scherer, VoxNet: A 3D Convolutional Neural Network for real-time object recognition intelligent robots and systems. pp. 922- 928 ,(2015) , 10.1109/IROS.2015.7353481
Yoshua Bengio, Jérôme Louradour, Ronan Collobert, Jason Weston, Curriculum learning Proceedings of the 26th Annual International Conference on Machine Learning - ICML '09. pp. 41- 48 ,(2009) , 10.1145/1553374.1553380
Yoshua Bengio, Saizheng Zhang, Aaron C. Courville, Anirudh Goyal Alias Parth Goyal, Alex M. Lamb, Ying Zhang, Professor Forcing: A New Algorithm for Training Recurrent Networks neural information processing systems. ,vol. 29, pp. 4601- 4609 ,(2016)
R. Qi Charles, Hao Su, Mo Kaichun, Leonidas J. Guibas, PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation computer vision and pattern recognition. pp. 77- 85 ,(2017) , 10.1109/CVPR.2017.16
Olga Diamanti, Leonidas J. Guibas, Ioannis Mitliagkas, Panos Achlioptas, Learning Representations and Generative Models for 3D Point Clouds international conference on machine learning. pp. 40- 49 ,(2018)
Vladlen Koltun, J. Zico Kolter, Shaojie Bai, An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling arXiv: Learning. ,(2018)
Yangyan Li, Baoquan Chen, Mingchao Sun, Wei Wu, Xinhan Di, Rui Bu, PointCNN: convolution on Χ -transformed points neural information processing systems. ,vol. 31, pp. 828- 838 ,(2018)
Frank Hutter, Ilya Loshchilov, Decoupled Weight Decay Regularization. international conference on learning representations. ,(2018)
Ahmad El Sallab, Mohamed Zahran, Ibrahim Sobh, Mahmoud Zidan, Sherif Abdelkarim, YOLO4D: A Spatio-temporal Approach for Real-time Multi-object Detection and Classification from LiDAR Point Clouds ,(2018)