monai.transforms.RandAffined

[agokli]

I am new to MONAI and trying to understand class monai.transforms.RandAffined. I have seen several tutorials using this specific transform. I still do not understand parameters. For an example, rand_affine = RandAffined(
keys=["image", "label"],
mode=("bilinear", "nearest"),
prob=1.0,
spatial_size=(300, 300, 50),
translate_range=(40, 40, 2),
rotate_range=(np.pi / 36, np.pi / 36, np.pi / 4),
scale_range=(0.15, 0.15, 0.15),
padding_mode="border",
)
(a) I am assuming that spatial_size is output batch size.
(b) How does rotate_range parameters get translated in rotational angles?
(c) How are rotate_range, translate_range and scale_range are correlated
(d) In general, how do you determine these parameters values? Any articles for providing such understanding?

[KumoLiu]

Hi @agokli, thanks for your interest here.
Hope you can find some information from the doc-string. The parameter selection often requires experimentation and fine-tuning depending on the specific application, perhaps you can try more experiments and check the output.

MONAI/monai/transforms/spatial/array.py

Line 2343 in cea80a6

rotate_range: angle range in radians. If element `i` is a pair of (min, max) values, then