DriveSim.md

Drive Simulator

Follow the instructions to train a generative model as described by Learning a Driving Simulator.

1. Download dataset

./get_data.sh

2. Start data server in the first terminal session

./server.py --batch 64

For maximum performance we recommend using tmux. screen is not good.
3) Train auto encoder in the second terminal

./train_generative_model.py autoencoder --batch 64

This will create two folders called outputs/results_autoencoder with model checkpoints and outputs/samples_autoencoder with samples from the generative model. In the sampled images odd columns are generated and even column images are target images.

Once the autoencoder model is trained you can stop the previous server and start training the transition model

4). Run server for transition model

./server.py --time 60 --batch 64

5. Train transition model

./train_generative_model.py transition --batch 64 --name transition

This will create two folders called outputs/results_transition with model checkpoints and outputs/samples_transition with samples from the generative model

Training logs will be saved to /tmp/logs/<model_name>. You can visualize logs using Tensorboard by typing tensorborad --logdir /tmp/logs/autoencoder or
tensorborad --logdir /tmp/logs/transition

6. Make a gif of the transition model

./view_generative_model.py transition --name transition

Here your job is to make the video look so great that it could be used train a steering angle model.