Mutli-Label Classification is a task where there are multiple labels to classify for a single image
- We change our targets from one-hot to Multi-hot encoding as there will be multiple true labels eg: [0 1 1 0 0 0 1 0]
- We change the loss function from
Cross-Entropy-Loss
toBinary-Cross-Entropy-Loss
because former is designed to classify only one true label whereas later allows multiple true labels to be classifed.