Python based Tinder automation script, created to solve mankind's most important need.
Includes: Tinder Data Creation Tool, Automation script.
I created the tool using the Tinder API, PyQt5 and some other libraries like pandas. The generated data was an actual representation of final data sample in this case. However, initially my data was about 500 images and highly imbalanced (Roughly 380 likes). So, I generated about 1500 images and it turns out I had a lot of dislikes later on in these images, ultimately balancing the dataset. Now I connected the labeled images to Azure's CustomVision platform because right off the bat I can use their transfer learning model (General) and get prediction API URL. I wrote the autotinder.py script to direct the image URL from the Tinder API's recommendation call to CustomVision's Prediction API call and got a like or dislike response. I then used the Tinder API to register the like/dislike call. But I was not done yet, this could be automated even more. Instead of me running the script manually every 12 hours or so, wouldn't it be great if the script ran automatically after every 12 hours (time taken to regenerate likes). I was able to achieve this through Azure's WebJobs. This was a little bit tricky as I had to bundle the required libraries with the script.
There were a lot of images where it was borderline like/dislike, on top of that group selfies made it even more problematic and the good old useless images of animals and memes had to be disliked. Their general Model doesn't specifically recognize faces, which in addition to the previously mentioned reasons explains the results.
- Azure WebJobs
- Azure CustomVision
- TinderAPI I got to know there is another cool API that you can use Pynder
- python 3.6
- requests
- robobrowser
- BeautifulSoup4
- lxml
- PyQt5
$ python main.py
$ python tindertool.py
Instructions
Train on VGG-Face Model to get better results.