Posted at 21:05h
in
dating
Thus, I accessed the Tinder API utilizing pynder
We wrote a software in which I could swipe through each profile, and cut each graphics to a "likes" folder or a "dislikes" folder. We invested never ending hours swiping and accumulated about 10,000 images.
One problem I noticed, was actually we swiped leftover approximately 80per cent for the pages. As a result, I got about 8000 in dislikes and 2000 in likes folder. That is a severely imbalanced dataset. Because You will find these types of couple of artwork for your loves folder, the date-ta miner will not be well trained to know what i love. It will just understand what We hate.
To fix this problem, I found files on the internet of individuals I found appealing. I then scraped these images and made use of all of them in my dataset.
What this API enables us to manage, are incorporate Tinder through my terminal screen rather than the app:
Given that We have the photographs, there are a number of problems. Some users have imagery with multiple friends. Some images become zoomed out. Some imagery become poor quality. It can hard to draw out information from these types of a top version of images.
To fix this dilemma, we put a Haars Cascade Classifier formula to draw out the confronts from images and then saved it.
The Algorithm failed to recognize the faces for around 70percent associated with the information. This is why, my dataset ended up being sliced into a dataset of 3,000 images.