Multi-Task Learning on Fate Grand Order with Keras
Published in
10 min readJan 21, 2019
Standard image classification tasks follow the trend of binary classification (cat v dog) or maybe multi-class classification (cat v dog v person) but many real world tasks like understanding an image or text usually require more than a single label to be useful. Out of this need to be more useful I think a lot of tasks can get framed as a multi-label in nature, essentially that a single sample has…