Compostion and phase fraction calculations added

Author: Alaukiksaxena

compositionspace/datautils.py

@@ -274,6 +274,8 @@ def get_big_slices(self):
             filestring = os.path.join(prefix, filestring)
             filestrings.append(filestring)
 
+            #pbar = tqdm(os.listdir(self.params["input_path"]), desc="Reading files")##
+            #for filename in pbar:
             hdf = h5py.File(filestring, "w")
             group1 = hdf.create_group("group_xyz_Da_spec")
             group1.attrs["columns"] = ["x","y","z","Da","spec"]
@@ -503,11 +505,14 @@ def calculate_voxel_composition(self, fileindex=0, outfilename="output_vox_ratio
             dic_ratios = {}
             for spec_name in spec_names:
                 dic_ratios["{}".format(spec_name)] = []
+            for spec_name in spec_names:
+                dic_ratios["{}_atoms".format(spec_name)] = []
 
             dic_ratios["Total_no"]=[]
             dic_ratios["file_name"]=[]
             dic_ratios["vox"] = []
 
+
             ratios = []
             f_count = 0
             pbar = tqdm(files, desc="Calculating voxel composition")
@@ -519,6 +524,7 @@ def calculate_voxel_composition(self, fileindex=0, outfilename="output_vox_ratio
                 for spec in (spec_names):
                     ratio = (len(np.argwhere(arr==spec)))/N_x
                     dic_ratios["{}".format(spec)].append(ratio)
+                    dic_ratios["{}_atoms".format(spec)].append(len(np.argwhere(arr==spec))) ## AS 27.10.2024 added for compostion calculations
 
                 dic_ratios["file_name"].append(filename)
                 dic_ratios["vox"].append(f_count)

compositionspace/postprocessing.py

@@ -123,7 +123,7 @@ def DBSCAN_clustering(self, voxel_centroid_phases_files, cluster_id,
             pointsToVTK(OutFile,x,y,z, data = {"label" : label}  )
 
         if save == True:
-            OutFile = os.path.join(self.params["output_path"], f"Output_DBSCAN_segmentation_phase{cluster_id}")
+            OutFile = os.path.join(self.params["output_path"], f"Output_DBSCAN_segmentation_phase{cluster_id}.h5")
             with h5py.File(OutFile, "w") as hdfw:
                 G = hdfw.create_group(f"{cluster_id}")
                 G.attrs["columns"] = Phase_columns

compositionspace/segmentation.py

@@ -169,8 +169,17 @@ def get_composition_cluster_files(self, vox_ratio_file, vox_file, n_components):
             group_name = list(list(hdfr.attrs.values())[1])
 
         ratios = pd.DataFrame(data=ratios, columns=ratios_columns) 
-        
-        X_train=ratios.drop(['Total_no','vox'], axis=1)
+
+        col_rm = []
+        for i in ratios_columns:
+            if "atoms" in i:
+                col_rm.append(i)
+        col_rm.append('Total_no')
+        col_rm.append('vox')
+
+        #X_train=ratios.drop(['Total_no','vox'], axis=1)
+        X_train=ratios.drop(col_rm, axis=1)
+
 
         gm = get_model(ml_params=ml_params)
         gm.fit(X_train)

tests/20230303_Workflow.ipynb

@@ -3,16 +3,16 @@ output_path: output
 n_big_slices: 10
 voxel_size: 2
 bics_clusters: 10
-n_phases: 2
+n_phases: 3
 ml_models:
    name: GaussianMixture 
    GaussianMixture:
-      n_components: 2
+      n_components: 3
       max_iter: 100000
       verbose: 0
    RandomForest:
       max_depth: 0
       n_estimators: 0
    DBScan:
       eps: 5
-      min_samples: 25
+      min_samples: 29

tests/experiment_params.yaml

@@ -0,0 +1,5 @@
+25.10.2024
+1. I added some calculation phase composition and error functionalities
+for nikita polins dataset(Fe doped Sm Co alloy)
+
+2. Interesesting to see that Fe and Cu are equally imporatnat for phase assignment