@ -120,7 +120,7 @@ You can use bounding boxes combined with probabilities to evaluate how accurate

In the example above, one bounding box indicated a predicted can of tomato paste at 78.3%. A second bounding box is slightly smaller, and is inside the first bounding box with a probability of 64.3%. Your code can check the bounding boxes, see they overlap completely, and ignore the lower probability as there is no way one can can be inside another.
In the example above, one bounding box indicated a predicted can of tomato paste at 78.3%. A second bounding box is slightly smaller, and is inside the first bounding box with a probability of 64.3%. Your code can check the bounding boxes, see they overlap completely, and ignore the lower probability as there is no way one can be inside another.
✅ Can you think of a situation where is it valid to detect one object inside another?