Add files via upload

add tools for rigid version

dynamight/inverse_deformations/compute_rigid_transforms.py

@@ -42,6 +42,7 @@ def compute_rigid_transforms(
     data_loader_threads: int = Option(4),
     pipeline_control=None,
     mask=None,
+    rigid: Optional[bool] = Option(False),
 ):
     forward_deformations_directory = output_directory / \
         'forward_deformations' / 'checkpoints'
@@ -72,6 +73,8 @@ def compute_rigid_transforms(
     decoder_half1.mask = None
     decoder_half2.mask = None
 
+    body = None
+
     n_points = decoder_half1.n_points
 
     points = decoder_half1.model_positions.detach().cpu()
@@ -94,25 +97,42 @@ def compute_rigid_transforms(
     encoder_half2.to(device)
 
     masked_points = []
-    for file in sorted(os.listdir(masks_directory)):
-
+    if rigid == False:
         try:
-            if file.startswith('mask'):
-                print('reading in:', file)
-                with mrcfile.open(masks_directory / file) as mrc:
-                    mask = torch.tensor(mrc.data).to(device)
-                    mask = mask.movedim(0, 2).movedim(0, 1)
-                    m_points, inds = maskpoints(
-                        decoder_half1.model_positions, decoder_half1.ampvar, mask, decoder_half1.box_size)
-                    print(m_points.shape)
-                    masked_points.append(m_points)
+            for file in sorted(os.listdir(masks_directory)):
+
+                try:
+                    if file.startswith('mask'):
+                        print('reading in:', file)
+                        with mrcfile.open(masks_directory / file) as mrc:
+                            mask = torch.tensor(mrc.data).to(device)
+                            mask = mask.movedim(0, 2).movedim(0, 1)
+                            m_points, inds = maskpoints(
+                                decoder_half1.model_positions, decoder_half1.ampvar, mask, decoder_half1.box_size)
+                            print(m_points.shape)
+                            masked_points.append(m_points)
+                except:
+                    print('No masks in this directory')
         except:
-            print('No masks in this directory')
-    print(len(masked_points))
+            for file in sorted(os.listdir(output_directory)):
+                print(file)
+                try:
+                    if 'mask' in file:
+                        print('reading in:', file)
+                        with mrcfile.open(output_directory / file) as mrc:
+                            mask = torch.tensor(mrc.data).to(device)
+                            mask = mask.movedim(0, 2).movedim(0, 1)
+                            m_points, inds = maskpoints(
+                                decoder_half1.model_positions, decoder_half1.ampvar, mask, decoder_half1.box_size)
+                            print(m_points.shape)
+                            masked_points.append(m_points)
+                except:
+                    print('No masks in this directory')
+
     latent_dim = encoder_half1.latent_dim
 
     star_file = starfile.read(refinement_star_file)
-    star_directory = output_directory / 'subsets'
+    star_directory = output_directory / 'body_starfiles'
     star_directory.mkdir(exist_ok=True, parents=True)
 
     relion_dataset = RelionDataset(
@@ -161,13 +181,20 @@ def compute_rigid_transforms(
         circular_mask_radius=diameter_ang / (2 * ang_pix),
         circular_mask_thickness=mask_soft_edge_width / ang_pix
     )
-
-    for i in range(len(masked_points)):
+    if rigid == True:
+        n_bodies = decoder_half1.n_bodies
+        masked_points = list(np.arange(n_bodies))
+    else:
+        n_bodies = len(masked_points)
+
+    for i in range(n_bodies):
+        if rigid == True:
+            body = i
+            print('Computing new star files for body ', body+1)
         current_star_file = star_file.copy()
         new_star = get_rotation_translation(
-            encoder_half1, decoder_half1, data_loader_half1, poses, data_preprocessor, [masked_points[i]], [current_star_file], half=1)
-        print(len(new_star))
+            encoder_half1, decoder_half1, data_loader_half1, poses, data_preprocessor, [masked_points[i]], [current_star_file], body, half=1)
         new_star = get_rotation_translation(
-            encoder_half2, decoder_half2, data_loader_half2, poses, data_preprocessor, [masked_points[i]], new_star, half=2)
+            encoder_half2, decoder_half2, data_loader_half2, poses, data_preprocessor, [masked_points[i]], new_star, body, half=2)
         starfile.write(new_star[0], star_directory /
-                       ('body_' + str(i+1) + '.star'))
+                       ('body_' + str(i+1) + '.star'), overwrite=True)