The model hasn't ever picked up or held a hammer. How is it ever supposed to recognize a hammer in an unfamiliar context (e.g., when given a new image that is n

The model hasn't ever picked up or held a hammer. How is it ever supposed to recognize a hammer in an unfamiliar context (e.g., when given a new image that is not representative of what it was trained against?). You can't train intuition.