Spaces:
Sleeping
Sleeping
File size: 2,035 Bytes
bf88397 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 | """Optional fixed surface filters in original OCT coordinates; no label inputs."""
from __future__ import annotations
import numpy as np
def apply_policy(positions, device, policy):
if not policy:
return positions
allowed = {
"median_b",
"median_a",
"gaussian_b_sigma",
"integer_devices",
"by_device",
}
if set(policy) - allowed:
raise ValueError("Unknown inference policy")
settings = {k: v for k, v in policy.items() if k != "by_device"}
if "by_device" in policy:
settings.update(policy["by_device"].get(device, {}))
if set(settings) - (allowed - {"by_device"}):
raise ValueError("Unknown device inference policy")
result = np.asarray(positions, dtype=np.float32)
if result.ndim != 3 or not np.isfinite(result).all():
raise ValueError("Invalid surface filter input")
for field, axis in [("median_b", 1), ("median_a", 2)]:
size = settings.get(field, 1)
if not isinstance(size, int) or size < 1 or size % 2 != 1:
raise ValueError("Median size must be a positive odd integer")
if size > 1:
from scipy.ndimage import median_filter
shape = [1, 1, 1]
shape[axis] = size
result = median_filter(result, size=shape, mode="nearest")
sigma = settings.get("gaussian_b_sigma", 0.0)
if not np.isfinite(sigma) or sigma < 0:
raise ValueError("Invalid smoothing sigma")
if sigma:
from scipy.ndimage import gaussian_filter1d
result = gaussian_filter1d(result, sigma, axis=1, mode="nearest")
integer_devices = settings.get("integer_devices", [])
if any(
d not in ["topcon_maestro2", "topcon_triton", "zeiss_cirrus"]
for d in integer_devices
):
raise ValueError(
"Integer policy is restricted to audited integer-coordinate families"
)
if device in integer_devices:
result = np.rint(result)
return result.astype(np.float32, copy=False)
|