Clarification on Euclidean Loss

[Guru Pradhan]

A very simple question, I am reading the documentation on Euclidean Loss http://caffe.berkeleyvision.org/doxygen/classcaffe_1_1EuclideanLossLayer.html#details and wanted to understand something. Supposing that the output of your CNN is a 2x2 image and your loss function is Euclidean (a regression task), how exactly is the loss computed? For example, if my testing net has a batch size of 2 with sample output as below:

Network Output = array([[[2, 3],

                                    [4, 5]],

                                   [[6, 7],

                                   [8, 9]]])

Target = array([[[2, 4],

                       [5, 3]],

                       [[6, 8],

                       [9, 9]]])

Will the loss be = (1/(2*2)) * [((2-2)**2 + (3-4)**2 + (4-5)**2 + (5-3)**2) + ((6-6)**2 + (7-8)**2 + (8-9)**2 + (9-9)**2)]

                       = (1/4) * [(0 + 1 + 1 + 4) + (0 + 1 + 1 + 0)]

                       = (1/4)* [6 + 2]

                       = 2?