The problem is, very few images have a direct mapping between dark/light and height/depth — if you have such an image, it will make a model as expected — if you don’t, you will be better served by:
If you have a suitable pixel image, then w/ a Pro license just use: