question for CNN example: Convolutional neural network for Regression

link for example:https://kr.mathworks.com/help/nnet/examples/train-a-convolutional-neural-network-for-regression.html

I'm using Convolutional neural network for Regression to predict 6 output(In example above the number of output is only 1, the rotation angle of figure).

Training result is pretty bad and i can't figure out why. please help me. I used 10200 data and divided it in ratio 0.7,0.15,0.15 for training, validation, test each.

This is the result.

And this is histogram of 6 outputs.

I think there is some relationship between these skewed output data and bad predicted value but i don't know exactly.