[WEEK 3-Country Classification Using House Photos]

Team Members: Meltem Tokgöz , Enes Furkan Çiğdem , Asma Aiouez

This week, we have completed the initial planning for our project. We have resized the 1800 pictures that had been collected. Moreover, we have performed The Augmentation method on our dataset. In the following part, we will be giving some information about the mentioned method.

When opting for deep learning methods, data size is an important factor to consider. The more data we have to train the more confident we are about the model’s ability to generalize. However, we realized that the amount of the dataset we obtained for each class is not enough. We made some research about the lack of data in the machine and deep learning and looked for ways that would help us increase the data size that we’ll use to feed to our ML classifiers and Augmentation method is the solution we found.

Basic augmentation techniques are as follows:

1.Flip : We can flip the images horizantally or vertically.

2. Rotation : We can rotate the image by some degree. However, if the image is a rectangle, dimensions of image won’t be preserved.

3.Scale : Outward and inward scaling can be done. Outward scaling is performed by increasing the size resulting in a larger sizes of the images. Whereas in inward scaling, the size of images is decreased and some assumptions about the image need to be done.

4.Zoom : This augmentation zooms in an image by an arbitrary amount. It helps the network invariant to size and also to predict well when part of the object is not in the image.

5. Translation : Translation process is just moving an image along X , Y or both directions. Objects which are in the image are located randomly.

By this manner, we have completed the data preprocession phase. In addition to all this, this week we have decided two of the algorithms that we will apply for the training part. In the next section, we provide some information about the to-be used algorithms:

1-Logistic Regression

Logistic Regression is a Machine Learning classification algorithm that is used to predict the probability of a categorical dependent variable. In logistic regression, the dependent variable is a binary variable that contains data coded as 1 (yes, success, etc.) or 0 (no, failure, etc.). In other words, the logistic regression model predicts P(Y=1) as a function of X.

For our project, we will be using multiclass logistic regression as we have 6 classes. Instead of y=0,1 we will expand our definition so that y=0,1,2,3,4,5. Basically we re-run binary classification multiple times, once for each class and for this algorithm we will resize our data to different sizes. For instance, 32x32 and 64x64 image size.

2-Single Layer Neural Network

A single-layer neural network represents the most simple form of neural network, in which there is only one layer of input nodes that send weighted inputs to a subsequent layer of receiving nodes, or in some cases, one receiving node. This single-layer design was part of the foundation for systems which have now become much more complex.

For the upcoming weeks, we are planning to perform CNN algorithm. However, we decided to apply Single Layer Neural Network algorithm for now.

In short, this week, data preprocessing part of the project has been dealt and research about the algorithms to employ has been done.