Abstract
本文: DeepTest
Task: a systematic testing tool for DNN-driven vehicles
Method:
- generated test cases with real-world changes like rain, fog, lighting conditions, etc.
- max the number of activated neurons
Github: https://github.com/ARiSE-Lab/deepTest
实验:
dataset: Udacity self-driving car challenge(Rambo, Chauffeur, Epoch)
效果:
- found thousands of erroneous behaviors under different realistic driving conditions