Add back the GDSII code

dflat/GDSII/assemble.py

@@ -0,0 +1,107 @@
+import cv2
+import numpy as np
+from tqdm.auto import tqdm
+import gdspy
+import time
+
+
+def assemble_nanocylinder_gds(
+    params,
+    mask,
+    cell_size,
+    block_size,
+    savepath,
+    gds_unit=1e-6,
+    gds_precision=1e-9,
+    marker_size=250e-6,
+    number_of_points=9,
+):
+    """Generate a GDS for nanocylinder metasurfaces.
+
+    Args:
+        params (float): Nanocylinder radius across the lens of shape [H, W, 1].
+        mask (int): Boolean mask of whether to write a shape or skip it of shape [H, W].
+        cell_size (list): Cell sizes holding the nanocylinder of [dy, dx].
+        block_size (list): Block sizes to repeat the nanocylinders of [dy', dx']. resize function is applied.
+        savepath (str): Path to save the gds file (including .gds extension).
+        gds_unit (flaot, optional): gdspy units. Defaults to 1e-6.
+        gds_precision (float, optional): gdspy precision. Defaults to 1e-9.
+        marker_size (float, optional): size of alignment markers. Defaults to 250e-6.
+        number_of_points (int, optional): Number of points to represent the shape. Defaults to 9.
+    """
+    assert len(cell_size) == 2
+    assert len(block_size) == 2
+    assert np.all(np.greater_equal(block_size, cell_size))
+    assert len(params.shape) == 3
+    assert len(mask.shape) == 2
+    assert mask.shape == params.shape[0:2]
+    assert params.shape[-1] == 1, "Shape dimension D should be equal to 1 (diameter)"
+
+    # upsample the params to match the target blocks
+    H, W, C = params.shape
+    scale_factor = np.array(block_size) / np.array(cell_size)
+    Hnew = np.rint(H * scale_factor[0]).astype(int)
+    Wnew = np.rint(W * scale_factor[1]).astype(int)
+
+    params_ = cv2.resize(
+        params,
+        (Wnew, Hnew),
+        interpolation=cv2.INTER_LINEAR,
+    )
+    params_ = np.expand_dims(params_, -1)
+    pshape = params_.shape
+    mask = cv2.resize(
+        np.expand_dims(mask, -1),
+        (Wnew, Hnew),
+        interpolation=cv2.INTER_LINEAR,
+    )
+    mask = mask.astype(int).astype(bool)
+
+    # Write to GDS
+    lib = gdspy.GdsLibrary(unit=gds_unit, precision=gds_precision)
+    cell = lib.new_cell("MAIN")
+    print("Writing metasurface shapes to GDS File")
+    start = time.time()
+    for yi in tqdm(range(pshape[0])):
+        for xi in range(pshape[1]):
+            if mask[yi, xi]:
+                xoffset = cell_size[1] * xi / gds_unit
+                yoffset = cell_size[0] * yi / gds_unit
+                radius = params_[yi, xi, 0] / gds_unit
+                circle = gdspy.Round((xoffset, yoffset), radius, number_of_points=9)
+                cell.add(circle)
+
+    ### Add some lens markers
+    hx = cell_size[1] * pshape[1] / gds_unit
+    hy = cell_size[0] * pshape[0] / gds_unit
+    ms = marker_size / gds_unit
+    cell_annot = lib.new_cell("TEXT")
+    add_marker_tag(cell_annot, ms, hx, hy)
+
+    top_cell = lib.new_cell("TOP_CELL")
+    # Reference cell1 and cell2 in the top-level cell
+    ref_cell1 = gdspy.CellReference(cell)
+    ref_cell2 = gdspy.CellReference(cell_annot)
+    top_cell.add(ref_cell1)
+    top_cell.add(ref_cell2)
+
+    lib.write_gds(savepath)
+    end = time.time()
+    print("Completed writing and saving metasurface GDS File: Time: ", end - start)
+
+    return
+
+
+def add_marker_tag(cell, ms, hx, hy):
+    mw = ms * 5 / 9
+    mh = ms / 2 + (21.10 / ms)  # Correction for the + sign center
+
+    cx = -mw / 2
+    cy = -mh + 2
+    # Add alignment markers at
+    cell.add(gdspy.Text("+", ms, (cx - ms / 2, cy - ms / 2)))  # lower left
+    cell.add(gdspy.Text("+", ms, (cx + hx + ms / 2, cy - ms / 2)))  # lower right
+    cell.add(gdspy.Text("+", ms, (cx + hx + ms / 2, cy + hx + ms / 2)))  # upper right
+    cell.add(gdspy.Text("+", ms, (cx - ms / 2, cy + hx + ms / 2)))  # upper left
+
+    return

dflat/metasurface/ckpt/Nanocylinders_TiO2_U180H600_Medium/config.yaml

@@ -1,7 +1,6 @@
 model:
   target: dflat.metasurface.optical_model.NeuralCells
   ckpt_path: metasurface/ckpt/Nanocylinders_TiO2_U180H600_Medium/model.ckpt
-  relative_ckpt: True
   params:
     trainable_model: False
     param_bounds:

requirements.txt

@@ -10,4 +10,6 @@ pyhank
 einops
 omegaconf
 tqdm
-pandas
+pandas
+opencv-python
+gdspy