soheil-mp/facial-keypoint-detection

Facial keypoint detection system takes in any image with faces, and predicts the location of 68 distinguishing keypoints on the face - Udacity project

facial-keypoint-detection's Issues

Do not use dropout on convolutional layers

Hei Soheil,

I would like to let you know that using dropout directy in CNN is not efficient at all. However, I will suggest that you use it between FC layers. Please read this article to have more information: https://towardsdatascience.com/dropout-on-convolutional-layers-is-weird-5c6ab14f19b2

Regards,

Axel

Why have you chosen mean=100 and std=50

scale keypoints to be centered around 0 with a range of [-1, 1]
# mean = 100, sqrt = 50, so, pts should be (pts - 100)/50
key_pts_copy = (key_pts_copy - 100)/50.0

In the above code why mean=100 and std=50,from where and how to calculate those values?