Udacity Machine Learning Nanodegree Capstone Project
Recommended setup for a simple Python script or web app/service:
- Python
- NumPy, SciPy, iPython
- TensorFlowโข
- Preprocess single.ipynb -- preproces the MNIST-like 32 * 32 format .mat data
- Train single.ipynb -- train a 5 CNN layer recognizing single digit
- Preprocess multi.ipynb -- preprocess the real image of SVHN dataset
- Train multi.ipynb -- built several models to get best sequence recognition accuracy
- Train Multi 2 - 64*64.ipynb -- building. The preprocessed image is 64 * 64
- Analysis.ipynb -- line chart for report
Capstone Project Report v2.pdf
Street View House Numbers (SVHN): A large-scale dataset of house numbers in Google Street View images.