RPN loss compute

[haigh1510]

Hi, I think that axis=(1,2,3) should be specified in np.sum

here

TensorFlow2.0-Examples/4-Object_Detection/RPN/train.py

Line 106 in f99fcef

score_loss = tf.reduce_sum(score_loss * foreground_background_mask, axis=[1,2,3]) / np.sum(foreground_background_mask)

and here

TensorFlow2.0-Examples/4-Object_Detection/RPN/train.py

Line 113 in f99fcef

boxes_loss = tf.reduce_sum(boxes_loss * foreground_mask, axis=[1,2,3]) / np.sum(foreground_mask)