Happy, Sad, or Angry?
#@title Importing the Lovelies %%capture ! [ -e /content ] && pip install -Uqq fastbook import fastbook fastbook.setup_book() !pip install nbdev from fastbook import * from fastai.vision.widgets import * import nbdev from fastai.vision import * from pathlib import Path import PIL !pip install -q jmd_imagescraper from jmd_imagescraper.core import * from jmd_imagescraper.imagecleaner import *
➢ Setting up the path where images will be saved
categories=['happy', 'sad', 'angry'] path = Path().cwd()/"faces" # assigns path to current working directory
➢ Downloading 400 images of each facial expression
It took a lot of experimenting with the best search strings for getting the most images of actual human faces showing the three emotions rather than getting primarily emojis showing the emotions. I did happen to get a model 100% accurate on discerning emoji emotions. But that was not nearly as interesting to me as the challenge of classifying the emotions from human image data.