Active Learning With Noisy Labelers for Improving Classification Accuracy of Connected Vehicles

摘要: Machine learning has emerged as a promising paradigm for enabling connected, automated vehicles to autonomously cruise the streets and react unexpected situations. Reacting such situations requires accurate classification uncommon events, which in turn depends on selection of large, diverse, high-quality training data. In fact, data available at vehicle (e.g., photos road signs) may be affected by errors or have different levels resolution freshness. To tackle this challenge, we propose an active framework that, leveraging information collected through onboard sensors well received from other vehicles, effectively deals with scarce noisy Given neighboring our solution: (i) selects can reliably generate data, (ii) obtains reliable subset add set trading off between two essential features, i.e., quality diversity. The results, obtained real-world datasets, demonstrate that significantly outperforms state-of-the-art solutions, providing high accuracy limited bandwidth requirement exchange vehicles.