diff --git a/DriveSim.md b/DriveSim.md index b566406..f684022 100644 --- a/DriveSim.md +++ b/DriveSim.md @@ -2,7 +2,7 @@ Follow the instructions to train a generative model as described by [Learning a Driving Simulator](http://arxiv.org/abs/1608.01230). -1) Download dataset +1) Download the dataset ``` ./get_data.sh ``` @@ -11,16 +11,16 @@ Follow the instructions to train a generative model as described by ```bash ./server.py --batch 64 ``` -For maximum performance we recommend using tmux. screen is not good. +For maximum performance, we recommend using tmux. Screen is not good. 3) Train auto encoder in the second terminal ```bash ./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. +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 +Once the autoencoder model is trained, you can stop the previous server and start training the transition model 4). Run server for transition model @@ -46,4 +46,4 @@ logs using Tensorboard by typing ``` -Here your job is to make the video look so great that it could be used train a steering angle model. +Here, your job is to make the video look so great that it could be used to train a steering angle model. diff --git a/README.md b/README.md index e654b67..502e33e 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ ## the comma.ai driving dataset -7 and a quarter hours of largely highway driving. Enough to train what we had in [Bloomberg](http://www.bloomberg.com/features/2015-george-hotz-self-driving-car/). +Contains 7 and a quarter hours of largely highway driving. Enough to train what we had in [Bloomberg](http://www.bloomberg.com/features/2015-george-hotz-self-driving-car/). ## Examples