Donkeycar has great documentation but we have a few convienace functions that need slightly different setup. Soooo I have litterally copied the docs from donkeycar and added the bits we need.
- ONLY If you dont have Conda
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash ./Miniconda3-latest-Linux-x86_64.sh
- cd to where you want to clone the dingocar repo, and clone.
cd my_repos
git clone https://github.com/tall-josh/dingocar.git
cd dingocar
git checkout master
- Create a dingocar environment
For tensorflow-cpu
conda env create -f install/envs/ubuntu-cpu.yml
For tensorflow-gpu
conda env create -f install/envs/ubuntu-gup.yml
- Install
conda activate dingo
pip install -e .[pc]
- Create your working dir
donkey createcar --path mycar
These are the steps to follow from gather data (manually drive) to training to driving with an Neural Network.
- Make sure your running an ssh server on your local machine. If not intall it.
sudo apt-get update
sudo apt-get install openssh-server
- Add your own public key to authorized keys
cat ~/.ssh/id_rsa.pub >> authorized_keys
- ssh to the pi with
-AR.Awill forward with you're local credentials. TheR 2050will forward a port from the Pi back to your local machine. This will allow our scrips on the Piscp(secure copy) files back to your loacl machine once you are done driving.
ssh -AR 2050:localhost:22 pi@roba1.local
# Now you're on the Pi
On the Pi
-
cd play/roba1/mycar -
Fill in the
info.json: Theinfo.jsonis used to help keep track of when a dataset was gathers and what type of data is in the tub that is gathered. I recomend you fill this is at the start of a labelling session (or whenever your track or track environment changes). A sampleinfo.jsonis below:
{
"count" : SET AUTOMATICALLY,
"date" : SET AUTOMATICALLY,
"tape" : "thick yellow",
"floor" : "timber",
"location" : "bill gates' house",
"centre line" : true,
# donkeycar: a python self driving library
[](https://travis-ci.org/autorope/donkeycar)
[](https://codecov.io/gh/autorope/donkeycar/branch/dev)
[](https://badge.fury.io/py/donkeycar)
[](https://img.shields.io/pypi/pyversions/donkeycar.svg)
Donkeycar is minimalist and modular self driving library for Python. It is
developed for hobbyists and students with a focus on allowing fast experimentation and easy
community contributions.
#### Quick Links
* [Donkeycar Updates & Examples](http://donkeycar.com)
* [Build instructions and Software documentation](http://docs.donkeycar.com)
* [Slack / Chat](https://donkey-slackin.herokuapp.com/)
#### Use Donkey if you want to:
* Make an RC car drive its self.
* Compete in self driving races like [DIY Robocars](http://diyrobocars.com)
* Experiment with autopilots, mapping computer vision and neural networks.
* Log sensor data. (images, user inputs, sensor readings)
* Drive your car via a web or game controller.
* Leverage community contributed driving data.
* Use existing CAD models for design upgrades.
### Get driving.
After building a Donkey2 you can turn on your car and go to http://localhost:8887 to drive.
### Modify your cars behavior.
The donkey car is controlled by running a sequence of events
```python
#Define a vehicle to take and record pictures 10 times per second.
import time
from donkeycar import Vehicle
from donkeycar.parts.cv import CvCam
from donkeycar.parts.datastore import TubWriter
V = Vehicle()
IMAGE_W = 160
IMAGE_H = 120
IMAGE_DEPTH = 3
#Add a camera part
cam = CvCam(image_w=IMAGE_W, image_h=IMAGE_H, image_d=IMAGE_DEPTH)
V.add(cam, outputs=['image'], threaded=True)
#warmup camera
while cam.run() is None:
time.sleep(1)
#add tub part to record images
tub = TubWriter(path='./dat',
inputs=['image'],
types=['image_array'])
V.add(tub, inputs=['image'], outputs=['num_records'])
#start the drive loop at 10 Hz
V.start(rate_hz=10)
See home page, docs or join the Slack channel to learn more.