Create preprocess_gt.py

preprocess_gt.py

@@ -0,0 +1,40 @@
+import os
+import numpy as np
+from plyfile import PlyData
+from tqdm import tqdm
+
+
+def preprocess(gt_path, save_path):
+    """Preprocess the CompleteScanNet dataset gt labels.
+
+    Args:
+        gt_path (str): Path to `CompleteScanNet_GT` directory.
+        save_path (str): Path where the preprocessed gt labels to be saved. The preprocessed labels 
+            is a ndarray each with shape (N, 7), N is for voxels number, 7 is for [x, y, z, r, g, b, label].
+    """
+    os.makedirs(save_path, exist_ok=True)
+    ply_paths = os.listdir(gt_path)
+    for p in tqdm(ply_paths, desc="Preprocessing gt labels: ", colour='green'):
+        pth = os.path.join(gt_path, p)
+        
+        ply_data = PlyData.read(pth)
+        vertex = ply_data['vertex']
+        new_xs = np.array(vertex['z'])
+        new_ys = np.array(vertex['x'])
+        new_zs = np.array(vertex['y'])
+        new_rs = np.array(vertex['red'])
+        new_gs = np.array(vertex['green'])
+        new_bs = np.array(vertex['blue'])
+        new_labels = np.array(vertex['label'])
+        voxels = np.stack([new_xs, new_ys, new_zs, new_rs, new_gs, new_bs, new_labels], axis=1)
+        
+        filename = os.path.join(save_path, p)
+        filename = filename.replace('ply', 'npy')
+        with open(filename, 'wb') as fp:
+            np.save(fp, voxels)
+    
+
+if __name__ == "__main__":
+    gt = os.getenv["COMPLETE_SCANNET_GT_PATH"]
+    preprocessed = os.getenv["COMPLETE_SCANNET_PREPROCESS_PATH"]
+    preprocess(gt, preprocessed)