This project is all about trying to identify images and if it's a meme or not. This is just of fun.
First, you need to install some dependencies first
The first step is to gather all the images. You can find a file named reddit_get_images.py in the module download_images. Let it run. It takes a while to download all the images. In average, it takes like 30-40 minutes.
Note: it only works once a day. This is because of how reddit works
After we got all the images, we need to have all the dependencies installed. After you have that, run a file named train_meme.py in the identify_memes module. It takes a while.
I have preselected some variables for you. Feel free to play with them as much as you want.
If you want more precision, then you have to download more images
Finally, if you trained correctly the images, feel free to evaluate what you just trained. In the identify_memes module there is a file named evauluate_memes.py. Run it.
The output is gonna have a lot of values, like true positives, false positives, true negatives and false negatives. Feel free to read this wikipedia article if you want more details.
Email: pduarte@nearsoft.com