improve metadata handling with ExposureSeries data model

pyproject.toml

@@ -43,6 +43,8 @@ dependencies = [
     "torch-grid-utils>=0.0.8",
     "torch-affine-utils",
     "torch-subpixel-crop",
+    "teamtomo-basemodel",
+    "pydantic",
 ]
 
 # https://peps.python.org/pep-0621/#dependencies-optional-dependencies

src/torch_tomogram/datamodel.py

@@ -0,0 +1,42 @@
+from typing import Annotated, Optional
+from datetime import datetime
+from teamtomo_basemodel import BaseModelTeamTomo, ExcludedTensor
+from pydantic import Field, AliasPath
+
+
+class RigidBodyAlignment(BaseModelTeamTomo):
+    """Rigid body alignment parameters
+
+    The alignment describes a forward projection model from the sample to
+    the record exposure in the detector.
+    """
+    # rotation parameters
+    z_rotation_degrees: float = Field(alias='tilt_axis_angle', default=0.0)
+    y_rotation_degrees: float = Field(alias='tilt_angle', default=0.0)
+    x_rotation_degrees: float = Field(alias='x_tilt', default=0.0)
+
+    # global image shifts
+    y_shift_angstrom: float = Field(default=0.0)
+    x_shift_angstrom: float = Field(default=0.0)
+
+
+class Optics(BaseModelTeamTomo):
+    defocus_u: float
+    defocus_v: float
+    # etc...
+
+
+class Exposure(BaseModelTeamTomo):
+    collection_time: Optional[datetime] = None
+    accumulated_dose: float = Field(default=0.0, gt=0.0)  # in e- per A^2
+
+    alignment: RigidBodyAlignment
+    optics: Optics
+
+
+class ExposureSeries(BaseModelTeamTomo):
+    pixel_spacing_angstrom: float = Field(gt=0.0)
+    image_stack: ExcludedTensor
+    exposure_list: list[Exposure] =(
+        Field(description="A list of metadata around tilt images")
+    )