C4W3 assignment (car detection with YOLO)

pass the tests,
but not sure how ft.gather() could extract “classes” even though tf.image.non_max_suppression() did not contain the “classes” .

This hint page did not seem to be helpful: tf.image.non_max_suppression  |  TensorFlow Core v2.7.0

You are right about nonmax suppression not containing classes.
We use the indices of the selected boxes from tf.image.non_max_suppression to pick the corresponding scores and classes. As part of def yolo_non_max_suppression exercise in the notebook, you get to implement this.