WEEK 2- Brand Recognition Using Machine Learning from Google Street View Images
Theme: Predicting the Brands
Team Members : Fırat Kaya, Hülya Şermin Karakaş
In the previous blog post, we were mentioned the explanation of the project and some related works. Until that time, in general, we ensured our dataset that we will use to train program and we worked on which tools we could use to make project.
The dataset we will use is COCO-Text: Dataset for Text Detection and Recognition [1] We decided to use that dataset because, on our researches, we saw that most of the related works that similar to our project used that dataset. In addition, this dataset contains optimum number of data that our computers able to work on it. For final testing, we are planning to use google street view dataset by reason of the aim of the project.
The tools we are planning to use are tensorflow tool with conventional neural network algorithm. [2] While we are trying to decide the tools, we especially considered ….. Let’s take a look at those tools.
We didn’t start to implement our algorithm to those tools, so we can not predict the results of the experiments yet. However, we will be started to try text recognition by using those tools in the near future. Hopefully, in the next week’s blog, we will be complete implementing the tensorflow tool to dataset task.
References
