Hi, Multi-Task learning is making a model learn to do multiple tasks like object detection, where the model has to learn to draw the bounding box and to classify the image, while Multi-label classification is just plain image classification problem, you ask the model to predict what’s in the image and that’s all it. What makes it “multi-label” is a photo may contain objects A and B, in this case, the model should output both A and B, whereas in ordinary image classification the model outputs only a single output either A or B.