Press enter or click to view image in full size
holds first place in the DeepTraficJS online challenge organized by MIT as part of the promotion of their 6.S094: Deep Learning for Self-Driving Cars course
He explains very well how he reached first place in this video - DeepTrafficJS solution:
Check the video description for resources that helped him achieve this great result.
As you’ll see in the video his intuition (and experience) contributed a lot in finding optimum parameters for the neural net, and a lot of training steps.
A MUST SEE video for anyone interested in deep learning and self driving cars.
I forked his solution here, for further integration in OSSDC:
To apply these methods in real world applications join us in the OSSDC Initiative.
For more news on Open Source DIY Self Driving Cars and Robotics follow me on Twitter @gtarobotics