I am a PhD student working on computational predictive coding. I really liked your work and would like to compare the PredNet architecture to a network I have implemented myself as a comparison. I was wondering how exactly you prepared the comma.ai dataset. I see that there are at times weird values (spikes) in their steering angle. Did you use some kind of threshold to get rid of very large numbers? If you have some code to share that would be great. I have attached a photo of a graph showing time vs steering angle for their 2016-04-21--14-48-08.h5 file.
Hi!
I am a PhD student working on computational predictive coding. I really liked your work and would like to compare the PredNet architecture to a network I have implemented myself as a comparison. I was wondering how exactly you prepared the comma.ai dataset. I see that there are at times weird values (spikes) in their steering angle. Did you use some kind of threshold to get rid of very large numbers? If you have some code to share that would be great. I have attached a photo of a graph showing time vs steering angle for their 2016-04-21--14-48-08.h5 file.
Thanks!