This is a repository to use AI to automate tinder stuff. AI part heavily based on Auto-Tinder, but it doesn't use the tinder API directly so it's more visual for the accompanying YouTube Video.
I do recommend just following Auto-Tinder if you don't want the visual aspect of swiping on a selenium instance because it complicates things a lot more. However, you can follow this messy guide to use this semi-messy code.
We need photos to train the AI on, so first off, you'll need a Tinder account linked to a google account, you'll need to set google_password and google_username as environment variables that correspond to your google credentials. Then you can run python extract_profiles.py , it will log you into google and prompt you please log into google hit enter after you finish 2 factor authentication.
The script will save profiles to data.json. Once you're satisfied with the amount of profiles you've extracted run python remove_dupes.py to make sure that you don't have duplicate profiles. You can also run python stats.py if you want to see some basic statistics on your dataset, however this isn't required.
Run python image_downloader.py this will download all the images in data.json
You'll now need to classify a like or dislike or not a person for all the photos you downloaded. Run python image_classifier.py , left click is a like, right click is a dislike, and middle mouse is if the photo isn't a person. (If you have non-people in your training data it might mess up the AI)
Run python prepare_data.py this will crop and convert images to gray-scale
Now you have to train the AI, MAKE SURE ALL THE DIRECTORIES EXIST THAT ARE REFERENCED IN THIS COMMAND (I made this mistake after 4 hours of training :( ), you can mess around with some of the arguments as to get a good model your training data is different from mine.
python retrain.py --bottleneck_dir=tf/training_data/bottlenecks --saved_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=25000 --testing_percentage=20 --learning_rate=0.0005
Follow the steps for environment variables in part 1, but run python use_model.py and hopefully it'll work for you!
There are some directories that don't auto generate and I forget what directories they are as they were defined in Auto-Tinder. I also probably forgot to include something here that's critical to getting this working.