n0shake / PepeClassifier

Machine Learning for the safety of the Pepe economy. 🐸

PepeClassifier

Created using Google's Inception model and the Codelab guide, this image classifier has a test accuracy of 86% in categorising your pepe into one of the following market categories : Laughing Pepe, Sad Pepe, Rare Pepe

Needs

Docker

How

Transfer learning allows you to apply the learning of a fully trained model for a new set of categories. The results are impressive for most applications, and the task does not require them GPUs.

Steps

1. Download Pepe images to train the network (I used about ~1000 images from google & rare-pepe dot com).
2. Create distinct categories (eg laughing pepe, crying pepe)
3. Retrain the network on these images
4. Test it against the new Pepes on the market

Credits

1. https://www.tensorflow.org/versions/r0.9/how_tos/image_retraining/index.html
2. https://github.com/xblaster

Stay cautious, brothers.