Large Scale Multimodal Data Capture, Evaluation and Maintenance Framework for Autonomous Driving Datasets

Nitheesh Lakshminarayana; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 0-0

Autonomous driving (AD) programs today primarily depend on one or the other form of supervised Deep Learning (DL) models for their behavioral success. However, these DL models are only as good as the data on which they are trained, and their success depends immensely on their training data. Hence it is imperative that we create datasets of good quality. However, the process of collecting this real-world driving data and the infrastructure needed to evaluate and manage this large data is commonly unspoken and is challenging. To address this issue, we have developed an open-source framework and infrastructure to capture, evaluate, and maintain such multi-sensor data. In this paper, we discuss the motive for this framework, a process for evaluation and quality analysis, insights on data storage, distribution and management for large multimodal data and the key lessons learned collecting and maintaining huge volumes of data from long driving distances.