-
Motion Selectivity of Neurons in Self-Driving Networks
-
Baladitya Yellapragada and Alexander Anderson and Stella X. Yu and Karl Zipser
-
European Conference on Computer Vision Workshop: Perceptual Organization in Computer Vision, Munich, Germany, 9 September 2018
-
Paper
|
Slides
-
Abstract
-
We investigated if optical flow filters were implicitly learned by a neural networktrained to drive a vehicle. The network was not trained to predict optical flow across theframes, but, through a series of controlled experiments, we show that optical flow filters are present in the network. However, this appears to be only the case for sideway flows more relevant for steering predictions. For motor throttle predictions, the network looks at the variance of the pixels over time rather than computing optical flows. In addition,the filters that are likely used for motor throttle predictions dominate primarily in the middle of the network.
-
Keywords
-
optical flow, motion selectivity, self-driving and autonomous driving, stereoscopic disparity
|