Matthew Tomei
ABSTRACT
Ensuring safety and reliability of autonomous vehicles requires good learning models which, in turn, require a large amount of real-world training data. Data produced by in-vehicle sensors (e.g., cameras, LIDARs, IMUs, etc.) can be used for training; however, both local storage and transmission of this sensor data to the cloud for subsequent use in training can be prohibitively expensive due to the staggering volume of data produced by these sensors, especially the cameras. In this paper, we perform the first exploration of techniques for reducing video frames in a way that the quality of training for autonomous vehicles is minimally affected. We particularly focus on utility aware data reduction schemes where the potential contribution of a video frame to enhancing the quality of learning (or utility) is explicitly considered during data reduction. Since actual utility of a video frame cannot be computed online, we use surrogate utility metrics to decide what video frames to keep for training and which ones to discard. Our results show that utility-aware data reduction schemes can reduce the amount of camera data required for training by as much as $16\times$ compared to random sampling for the same quality of learning (in terms of IoU).
- Chen, L.-C., Papandreou, G., Kokkinos, I., Murphy, K., and Yuille, A. L. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs. arXiv:1606.00915 [cs] (June 2016). arXiv: 1606.00915.Google Scholar
- Chen, L.-C., Papandreou, G., Schroff, F., and Adam, H. Rethinking AtrousConvolution for Semantic Image Segmentation. arXiv:1706.05587 [cs] (June 2017). arXiv: 1706.05587.Google Scholar
- Denker, J. S., and LeCun, Y. Transforming Neural-Net Output Levels to Probability Distributions. In Advances in Neural Information Processing Systems 3, R. P. Lippmann, J. E.Moody, andD. S. Touretzky, Eds. Morgan- Kaufmann, 1991, pp. 853--859.Google Scholar
- Everingham, M., Van Gool, L., Williams, C. K. I., Winn, J., and Zisserman, A. The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results. http://www.pascalnetwork. org/challenges/VOC/voc2012/workshop/index.html.Google Scholar
- Gal, Y., and Ghahramani, Z. Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. arXiv:1506.02142 [cs, stat] (June 2015). arXiv: 1506.02142.Google Scholar
- Geiger, A., Lenz, P., and Urtasun, R. Are we ready for autonomous driving? the kitti vision benchmark suite. In Proc. CVPR (2012).Google ScholarCross Ref
- Harris, D. Baidu's chief scientist explains why computers won't take over the world just yet, Sept. 2015.Google Scholar
- Kendall, A., Badrinarayanan, V., and Cipolla, R. Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding. arXiv:1511.02680 [cs] (Nov. 2015). arXiv: 1511.02680.Google Scholar
- Pedregosa, F., Varoqaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research 12 (2011), 2825--2830.Google ScholarDigital Library
- Real, R., and Vargas, J. M. The Probabilistic Basis of Jaccard's Index of Similarity. Systematic Biology 45, 3 (Sept. 1996), 380--385.Google Scholar
- Winter, K. For self-driving cars, thereâ's big meaning behind one big number: 4 terabytes. "https://newsroom.intel.com/editorials/ self-driving-cars-big-meaning-behind-one-number-4-terabytes/".Google Scholar
- Xfinity. What is the median usage of people on your network today?, June 2018.Google Scholar
Index Terms
- Sensor Training Data Reduction for Autonomous Vehicles
Recommendations
Autonomous Vehicles Empty Cruising Impact on Parking Dynamics
IWCTS '23: Proceedings of the 16th ACM SIGSPATIAL International Workshop on Computational Transportation ScienceAs development and utilization of autonomous vehicles is gaining momentum, it is expected to be decades of mixed environment where autonomous and human-driven vehicles will co-exist. Autonomous vehicles' ability to perform empty cruising can impact the ...
The Exploration of Autonomous Vehicle Driving Styles: Preferred Longitudinal, Lateral, and Vertical Accelerations
Automotive'UI 16: Proceedings of the 8th International Conference on Automotive User Interfaces and Interactive Vehicular ApplicationsThis paper describes a new approach in exploring preferred driving styles for autonomous vehicles through simulation of autonomous driving in real road conditions. A Wizard experiment with an equipped car was conducted to investigate the preferences of ...
Statistical accident analysis supporting the control of autonomous vehicles
Autonomous vehicles offer the potential to drastically decrease the number and severity of road accidents. Most accidents occur due to human inattention or wrong decisions, whose factors can be eliminated by autonomous vehicles. However, not all ...
Comments