In this project, I implemented Nvidia's deep learning methodologies and built a model which can take in the data of right, left, and central camera images from the car, preprocess it, and use it to drive a car autnomously without a driver.
I used Python's frameworks and libraries such as numpy, tensorflow (by Google), Keras and openCV to build the model. Furthermore, I used socketio to establish the connection of our model with the udacity simulator on localhost and the car drove autonomously.
The last results of this model showed a 92.47 % of efficiency in detecting the turns. I am still studying about the hypertuning and optimizing techniques.
Feel free to contribute and connect with me on similar projects.
Resources -
-
Nvidia's whitepaper - https://images.nvidia.com/content/tegra/automotive/images/2016/solutions/pdf/end-to-end-dl-using-px.pdf
-
OpenCV https://docs.opencv.org/4.x/
-
Tensorflow - https://www.tensorflow.org/api_docs
Email: [email protected]