Data Scraping (انتشار: 1402/07/13)

One of the ways to access the data set so that we can optimize the model method is Data Scraping. The data scraping technique allows us to collect the desired data from the Internet and various websites. In the Inchiyeh project, we need a large number of labeled images in different classes to get the best model result.

To start scraping, we need to search and find some websites that have labeled images. Maybe the images are not clearly and accurately labeled, but to some extent the image label is close to its classification can be used to start our work.

This image from the DigiKala website shows the products of clothing, men's clothing and shoes. You can save the images of this section and assign the three labels mentioned to the target. Depending on how many levels our classification is going to be, you can assign different labels and targets to the images. So far, one thing is clear and that is that we need a very high volume of samples to cover a large number of classes. First of all, we want to get our category list and collect related images based on the list we have. A good thing we want to do is to provide the datasets we collected and labeled for free to those interested in this work. We hope it will be useful for you.

داریوش مهدی پور یقینی