Add optional target physical-range mapping: scripts/target_mapping.py

scripts/target_mapping.py

@@ -0,0 +1,50 @@
+"""Utilities for mapping normalized target modalities to physical ranges."""
+
+from __future__ import annotations
+
+import numpy as np
+
+
+TARGET_CHANNELS = ["D", "Delta", "eta", "theta", "psi", "R"]
+
+
+def normalized_modalities_to_physical(target, channel_axis=0, clip=False):
+    """Map normalized grayscale modalities to nominal physical ranges.
+
+    Use this only for targets encoded as
+    ``png_uint8_normalized_to_float32_0_1``. If a split already stores physical
+    Lu-Chipman values, do not apply this conversion again.
+    """
+    target = np.asarray(target, dtype=np.float32)
+    values = np.moveaxis(target, channel_axis, 0)
+    if values.shape[0] != 6:
+        raise ValueError(f"Expected 6 target channels, got shape {target.shape}")
+
+    g = np.clip(values, 0.0, 1.0) if clip else values
+    physical = np.empty_like(g, dtype=np.float32)
+    physical[0] = g[0]
+    physical[1] = g[1]
+    physical[2] = np.pi * g[2]
+    physical[3] = np.pi * (g[3] - 0.5)
+    physical[4] = np.pi * (g[4] - 0.5)
+    physical[5] = np.pi * g[5]
+    return np.moveaxis(physical, 0, channel_axis)
+
+
+def physical_modalities_to_normalized(target, channel_axis=0, clip=False):
+    """Map physical target modalities to normalized grayscale ranges."""
+    target = np.asarray(target, dtype=np.float32)
+    values = np.moveaxis(target, channel_axis, 0)
+    if values.shape[0] != 6:
+        raise ValueError(f"Expected 6 target channels, got shape {target.shape}")
+
+    normalized = np.empty_like(values, dtype=np.float32)
+    normalized[0] = values[0]
+    normalized[1] = values[1]
+    normalized[2] = values[2] / np.pi
+    normalized[3] = values[3] / np.pi + 0.5
+    normalized[4] = values[4] / np.pi + 0.5
+    normalized[5] = values[5] / np.pi
+    if clip:
+        normalized = np.clip(normalized, 0.0, 1.0)
+    return np.moveaxis(normalized, 0, channel_axis)