Build a Digit Classifier
- The purpose of this notebook is to build a Digit Image classifier the Street View Housing Number dataset.
- Details of the problem statement , data set , sample input , summary of the code/solution and final result of the project are listed in the sections to follow.
A Recognising multi-digit numbers in photographs captured at street level is an important component of modern-day map making. A classic example of a corpus of such street-level photographs is Google’s Street View imagery composed of hundreds of millions of geo-located 360-degree panoramic images.
The SVHN is a real-world image dataset for developing machine learning and object recognition algorithms with the minimal requirement on data formatting but comes from a significantly harder, unsolved, real-world problem (recognising digits and numbers in natural scene images). SVHN is obtained from house numbers in Google Street View images. Where the labels for each of this image are the prominent number in that image i.e. 2,6,7 and 4 respectively. The dataset has been provided in the form of h5py files.
Autonomous Vehicles
The code aims at building a digit(Image) Classifier classifier
- We begin by doing an Exploratory Data analyses and Visualisation/viewing of the images
- We then do the required pre-processing of the data to prepare it so it can be fed into a Neural Network which involves converting target variable into a one hot vector and normalising the input data
- We then begin the process of building a Neural Network model
- Next we start "tuning" the model in terms of no of layers,adding L2 regularization parameter,adding batch normalisation layers & Drop out and for each of these cases , we capture the test data accuracy
- Finally we compare the test/validation accuracy for all the various models built above and choose the best contender
- Refer python worksheet Project_ClassificationUsingNeuralNetworks_MultiDigitNumbers.ipynb for the solution
