fix data loading, formating bugs and dataset path (#1913)

configs/_base_/datasets/waymoD5-3d-3class.py

@@ -88,7 +88,8 @@ train_dataloader = dict(
             type=dataset_type,
             data_root=data_root,
             ann_file='waymo_infos_train.pkl',
-            data_prefix=dict(pts='training/velodyne'),
+            data_prefix=dict(
+                pts='training/velodyne', sweeps='training/velodyne'),
             pipeline=train_pipeline,
             modality=input_modality,
             test_mode=False,
@@ -107,7 +108,7 @@ val_dataloader = dict(
     dataset=dict(
         type=dataset_type,
         data_root=data_root,
-        data_prefix=dict(pts='training/velodyne'),
+        data_prefix=dict(pts='training/velodyne', sweeps='training/velodyne'),
         ann_file='waymo_infos_val.pkl',
         pipeline=eval_pipeline,
         modality=input_modality,
@@ -124,7 +125,7 @@ test_dataloader = dict(
     dataset=dict(
         type=dataset_type,
         data_root=data_root,
-        data_prefix=dict(pts='training/velodyne'),
+        data_prefix=dict(pts='training/velodyne', sweeps='training/velodyne'),
         ann_file='waymo_infos_val.pkl',
         pipeline=eval_pipeline,
         modality=input_modality,

mmdet3d/datasets/transforms/formating.py

@@ -63,6 +63,7 @@ class Pack3DDetInputs(BaseTransform):
 
     def __init__(
         self,
+        keys: tuple,
         meta_keys: tuple = ('img_path', 'ori_shape', 'img_shape', 'lidar2img',
                             'depth2img', 'cam2img', 'pad_shape',
                             'scale_factor', 'flip', 'pcd_horizontal_flip',
@@ -74,7 +75,8 @@ class Pack3DDetInputs(BaseTransform):
                             'affine_aug', 'sweep_img_metas', 'ori_cam2img',
                             'cam2global', 'crop_offset', 'img_crop_offset',
                             'resize_img_shape', 'lidar2cam', 'ori_lidar2img',
-                            'num_ref_frames', 'num_views', 'ego2global')) -> None:
+                            'num_ref_frames', 'num_views', 'ego2global')
+    ) -> None:
         self.keys = keys
         self.meta_keys = meta_keys
 

mmdet3d/datasets/transforms/loading.py

@@ -165,7 +165,7 @@ class LoadMultiViewImageFromFiles(BaseTransform):
         results['ori_cam2img'] = copy.deepcopy(results['cam2img'])
 
         if self.file_client is None:
-            self.file_client = mmcv.FileClient(**self.file_client_args)
+            self.file_client = mmengine.FileClient(**self.file_client_args)
 
         # img is of shape (h, w, c, num_views)
         # h and w can be different for different views

mmdet3d/datasets/transforms/transforms_3d.py

@@ -2245,7 +2245,7 @@ class PhotoMetricDistortion3D(PhotoMetricDistortion):
 
 
 @TRANSFORMS.register_module()
-class MultiViewWrapper(object):
+class MultiViewWrapper(BaseTransform):
     """Wrap transformation from single-view into multi-view.
 
     The wrapper processes the images from multi-view one by one. For each
@@ -2290,7 +2290,7 @@ class MultiViewWrapper(object):
                      'scale', 'scale_factor', 'crop_size', 'img_crop_offset',
                      'flip', 'flip_direction', 'photometric_param'
                  ]):
-        self.transform = Compose(transforms)
+        self.transforms = Compose(transforms)
         self.override_aug_config = override_aug_config
         self.collected_keys = collected_keys
         self.process_fields = process_fields
@@ -2325,7 +2325,7 @@ class MultiViewWrapper(object):
             for key in self.process_fields:
                 if key in input_dict:
                     process_dict[key] = input_dict[key][img_id]
-            process_dict = self.transform(process_dict)
+            process_dict = self.transforms(process_dict)
             # store the randomness variable in transformation.
             prev_process_dict = process_dict