realitycapture nerfacto variable resolution (#1651)

* allow periods within filename, not only for file extension

* don't assume all images are the same resolution

* black reformatting

* Move variable resolution collate logic into vanilla datamanager

* good catch pylint

* now that vanilla supports variable resolution, remove VariableResDataManager

* remove comment

* CONSOLE.print

* avoid `IndexError: tensors used as indices must be long, byte or bool tensors`

* use large dataset parameters for phototourism

nerfstudio/process_data/realitycapture_utils.py

@@ -59,17 +59,6 @@ def realitycapture_to_json(
             for column, value in row.items():
                 cameras.setdefault(column, []).append(value)
 
-    for name in cameras["#name"]:
-        camera_label = name.rpartition(".")[0]
-        if camera_label in image_filename_map:
-            img = np.array(Image.open(output_dir / image_filename_map[camera_label]))
-            break
-
-    height, width, _ = img.shape
-
-    data["h"] = int(height)
-    data["w"] = int(width)
-
     missing_image_data = 0
 
     for i, name in enumerate(cameras["#name"]):
@@ -79,7 +68,12 @@ def realitycapture_to_json(
                 CONSOLE.print(f"Missing image for camera data {basename}, Skipping")
             missing_image_data += 1
             continue
+
         frame = {}
+        img = np.array(Image.open(output_dir / image_filename_map[basename]))
+        height, width, _ = img.shape
+        frame["h"] = int(height)
+        frame["w"] = int(width)
         frame["file_path"] = image_filename_map[basename].as_posix()
         frame["fl_x"] = float(cameras["f"][i]) * max(width, height) / 36
         frame["fl_y"] = float(cameras["f"][i]) * max(width, height) / 36