[Week — 4] Buildings In The City

Team Members: Ahmet Faruk BAYRAK, Mehmet Harun DOĞANAY, Mustafa DANYILDIZ

In the previous week we gathered the dataset and explained our project.

In this week, our main goal was determine the methodology of our project. We searched related works and looked for their methods and what kind of features they used in projects.

Also we wrote progress report about our project in this week. While writing the report, we made huge investigations and it helped us a lot. Examining the properties of buildings from paintings is not a very studied subject. There is not much related work like standard scene classification. In the related works and projects, in general Convolutional Neural Network or Support Vector Machine were used as a method. But Convolutional Neural Network was used more than Support Vector Machine. CNNs are often used in image recognition systems because of CNN has lower error rate and learning process is faster than others. So we decided to try both of them and then select the efficient one for our project.

In this week we did regulate the dataset also. We looked for the images and removed the irrelevant ones. We removed them for increasing the accurate and successful train. After removing our dataset looks inadequate. We will try that dataset if it is inadequate, we are going to add new photos to our dataset.

Other problem that we are going to solve is standardize our photos. We will try some methods according to related works and our investigations.

In the next week, we are going to start coding our project, try methods as mentioned above. We were good in theoretical part let’s see in the practice. See you in the next week!