A Convolutional Neural Network trained to classify pictures of men/women.
Following the fast.ai course on Neural Networks, I didn't want to simply reproduce the notebooks with the same results, so I tried to apply my knowledges to another problem.
Classifying men/women tend to be more tedious, because humans can be either very similar, or very different, in a physical way.
I've created this dataset on my own, collecting images from Google Images. It's manually collected and cleansed dataset containing 3.354 pictures (jpg) of men (1414 files) and women (1940 files). There should be close to none mislabeled files left.
Download the cleansed dataset from Kaggle @ https://www.kaggle.com/playlist/men-women-classification
You can also download the raw dataset from Google, following the instructions in the notebook
- fastai 1.0.51
- pandas
- numpy
- pytorch 1.0.1 OR pytorch-cpu 1.0.1