does similarity transform matrix depend on the image size?

I am trying to align image with different size (182x182) but I am not getting good results. The alignment looks good if I pass the default image_size (112x112). So, I am wondering if the `

  src = np.array([
  [30.2946, 51.6963],
  [65.5318, 51.5014],
  [48.0252, 71.7366],
  [33.5493, 92.3655],
  [62.7299, 92.2041] ], dtype=np.float32 )

if image_size[1]==112:
  src[:,0] += 8.0`

matrix given is for 112x112 images only,

[vkasojhaa]

@mlourencoeb Thank you, rescaling the values like you told resolved the issue.

[mlourencoeb]

Not really. After that line where the shift is applied all you needed to do is:

src = src * 182.0 / 112