This Tinder bot will learn your interests and automate Tinder swiping game using its API and Tensorflow.

• Python v3.10.1
• Tensorflow v2.8.0rc1
• All required python packages with versions are listed in requirements.txt. Simply run pip install -r requirements.txt to install all of them at once
• Optional (to utilize GPU acceleration):
  • CUDA-compatible NVidia GPU (Developed and tested on GeForce RTX 2080 Super Max-Q)
  • CUDA v11.2
  • CUDNN v8.2.1

First and foremost you'll need to acquire a Tinder token that will authenticate your account. To do so, in a browser open Developer tools (F12 in Firefox). Then, in the tab with Dev Tools go to https://tinder.com/app/recs. In Dev Tools, under the Network tab, search for a GET request to api.gotinder.com. In Headers, it should contain X-Auth-Token that looks something like 943e7695-c8eb-415a-524d-67783caafc65. Copy this token and paste it in config.py as TINDER_TOKEN value.

Just run py bot.py. For as long as you don't run out of the Likes limit on Tinder, it will be pulling more profiles filtered by preferences in your Tinder profile settings. Each profile's photos will be evaluated based on the given model to calculate the profile's score. If the score passes the set threshold, Bot will like the profile. Otherwise, it will pass it and go to the next profile. As of now, Tinder does not provide a strict Dislike function, all passed profiles will return to the pool of profiles from which they will be pulled again later.

1. Put a large number of images into the .images/unclassified folder.
2. Run py image_classifier.py which will pull those images one by one and let you mark them as positive (right arrow) or negative (left arrow). Bottom arrow will completely remove the image. It will sort them into ./images/classified/positive and ./images/classified/negative folders.
   Note: new window with images might require you to manually select it and press Tab a couple of times for the Next image button to get in focus for arrow keys to work properly.
3. Then run py prepare_data.py which will try and find a person on classified images, and grayscale the image for the model to work with.
4. After all images are classified and processed by those two scripts, run py retrain.py --bottleneck_dir=tf/training_data/bottlenecks --model_dir=tf/training_data/inception --summaries_dir=tf/training_data/summaries/basic --output_graph=tf/training_output/retrained_graph.pb --output_labels=tf/training_output/retrained_labels.txt --image_dir=./images/classified --how_many_training_steps=50000 --testing_percentage=20 --learning_rate=0.001 Note: parameters like --how_many_training_steps=50000 --testing_percentage=20 --learning_rate=0.001 might need to be adjsuted depending on your experience with precision results.