![]() ![]() Two separate datasets were used for steering and throttle. ![]() Layers were trained to create the full model. After this, a second regression head was added and only the remaining regression The model was initially only trained using the steering command to learn features in theĬonvolution layers. However, this network was modified to add two regression heads, one for each command. The architecture was loosely modelled after PilotNet by NVIDIA This model was trained using ~80,000 images with corresponding steering and throttle commands. The aim was to make a slightly more realistic GTA5 self-driving car.īy utilizing imitation learning the model has learned to stay in lane Images are at 2x speed to show more driving in a shorter time span. GTA5-Self-Driving-car Driving Unfamiliar Roads ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |