Udacity Self Driving Car Nanodegree - Project #2
2017/7/10
This project processes a German road sign dataset and trains a TensorFlow neural network to classify each sign image. The network architecture uses convolution and fully connected layers to achieve ~95% prediction accuracy. Five additional sign images found by Google Maps are also processed to demonstrate an example of real world application and investigate its accuracy and performance.
For more details about the results of this activity, see the project writeup document.
| File | Description |
|---|---|
| Traffic_Sign_Classifier.ipynbs | IPython notebook with all project code |
| report.html | Exported HTML notebook with saved results |
| writeup_P2.md | The project writeup explaining the results |
| /web_signs/websign[01~05].jpg | Five new sign images downloaded from the web |
The original Udacity project repository is here.
-
Set up Udacity Self-Driving Car Term 1 Starter Kit environment (Python 3, NumPy, Matplotlib, OpenCV, TensorFlow)
-
Open the IPython notebook "Traffic_Sign_Classifier.ipynbs" using Jupyter, and execute all cells.