[Feature] File I/O migration and reconstruction (#2319)

* refactor leio

* update mmengine and mmcv version

* update

* update docs

* update version

configs/ssn/ssn_hv_secfpn_sbn-all_16xb2-2x_lyft-3d.py

@@ -11,10 +11,19 @@ class_names = [
     'bicycle', 'motorcycle', 'pedestrian', 'animal', 'car',
     'emergency_vehicle', 'bus', 'other_vehicle', 'truck'
 ]
+backend_args = None
 
 train_pipeline = [
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=5, use_dim=5),
-    dict(type='LoadPointsFromMultiSweeps', sweeps_num=10),
+    dict(
+        type='LoadPointsFromFile',
+        coord_type='LIDAR',
+        load_dim=5,
+        use_dim=5,
+        backend_args=backend_args),
+    dict(
+        type='LoadPointsFromMultiSweeps',
+        sweeps_num=10,
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -34,8 +43,16 @@ train_pipeline = [
         keys=['points', 'gt_bboxes_3d', 'gt_labels_3d'])
 ]
 test_pipeline = [
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=5, use_dim=5),
-    dict(type='LoadPointsFromMultiSweeps', sweeps_num=10),
+    dict(
+        type='LoadPointsFromFile',
+        coord_type='LIDAR',
+        load_dim=5,
+        use_dim=5,
+        backend_args=backend_args),
+    dict(
+        type='LoadPointsFromMultiSweeps',
+        sweeps_num=10,
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),

configs/ssn/ssn_hv_secfpn_sbn-all_16xb2-2x_nus-3d.py

@@ -11,10 +11,19 @@ class_names = [
     'bicycle', 'motorcycle', 'pedestrian', 'traffic_cone', 'barrier', 'car',
     'truck', 'trailer', 'bus', 'construction_vehicle'
 ]
+backend_args = None
 
 train_pipeline = [
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=5, use_dim=5),
-    dict(type='LoadPointsFromMultiSweeps', sweeps_num=10),
+    dict(
+        type='LoadPointsFromFile',
+        coord_type='LIDAR',
+        load_dim=5,
+        use_dim=5,
+        backend_args=backend_args),
+    dict(
+        type='LoadPointsFromMultiSweeps',
+        sweeps_num=10,
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -34,8 +43,16 @@ train_pipeline = [
         keys=['points', 'gt_bboxes_3d', 'gt_labels_3d'])
 ]
 test_pipeline = [
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=5, use_dim=5),
-    dict(type='LoadPointsFromMultiSweeps', sweeps_num=10),
+    dict(
+        type='LoadPointsFromFile',
+        coord_type='LIDAR',
+        load_dim=5,
+        use_dim=5,
+        backend_args=backend_args),
+    dict(
+        type='LoadPointsFromMultiSweeps',
+        sweeps_num=10,
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),

docs/en/notes/faq.md

@@ -10,7 +10,7 @@ We list some potential troubles encountered by users and developers, along with
 
   | MMDetection3D version |     MMEngine version     |      MMCV version       |   MMDetection version    |
   | --------------------- | :----------------------: | :---------------------: | :----------------------: |
-  | dev-1.x               | mmengine>=0.4.0, \<1.0.0 | mmcv>=2.0.0rc4, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
+  | dev-1.x               | mmengine>=0.6.0, \<1.0.0 | mmcv>=2.0.0rc4, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc3             | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc2             | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc1             | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |

docs/en/user_guides/backends_support.md

@@ -8,7 +8,7 @@ We support loading data and generated annotation info files (pkl and json) from
 
 ```python
 # set file client backends as Ceph
-file_client_args = dict(
+backend_args = dict(
     backend='petrel',
     path_mapping=dict({
         './data/nuscenes/':
@@ -30,15 +30,15 @@ db_sampler = dict(
         coord_type='LIDAR',
         load_dim=4,
         use_dim=4,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     # set file client for data base sampler to load db info file
-    file_client_args=file_client_args)
+    backend_args=backend_args)
 
 train_pipeline = [
     # set file client for loading training data
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, file_client_args=file_client_args),
+    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, backend_args=backend_args),
     # set file client for loading training data annotations
-    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True, file_client_args=file_client_args),
+    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True, backend_args=backend_args),
     dict(type='ObjectSample', db_sampler=db_sampler),
     dict(
         type='ObjectNoise',
@@ -59,7 +59,7 @@ train_pipeline = [
 ]
 test_pipeline = [
     # set file client for loading validation/testing data
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, file_client_args=file_client_args),
+    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),
@@ -87,11 +87,11 @@ data = dict(
     train=dict(
         type='RepeatDataset',
         times=2,
-        dataset=dict(pipeline=train_pipeline, classes=class_names, file_client_args=file_client_args)),
+        dataset=dict(pipeline=train_pipeline, classes=class_names, backend_args=backend_args)),
     # set file client for loading validation info files (.pkl)
-    val=dict(pipeline=test_pipeline, classes=class_names,file_client_args=file_client_args),
+    val=dict(pipeline=test_pipeline, classes=class_names,backend_args=backend_args),
     # set file client for loading testing info files (.pkl)
-    test=dict(pipeline=test_pipeline, classes=class_names, file_client_args=file_client_args))
+    test=dict(pipeline=test_pipeline, classes=class_names, backend_args=backend_args))
 ```
 
 ## Load pretrained model from Ceph

docs/en/user_guides/config.md

@@ -466,11 +466,11 @@ train_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -491,11 +491,11 @@ test_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),

docs/en/user_guides/data_pipeline.md

@@ -28,11 +28,11 @@ train_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -52,11 +52,11 @@ test_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug',
         img_scale=(1333, 800),
@@ -176,11 +176,11 @@ For each operation, we list the related dict fields that are added/updated/remov
            type='LoadPointsFromFile',
            load_dim=5,
            use_dim=5,
-           file_client_args=file_client_args),
+           backend_args=backend_args),
        dict(
            type='LoadPointsFromMultiSweeps',
            sweeps_num=10,
-           file_client_args=file_client_args),
+           backend_args=backend_args),
        dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
        dict(
            type='GlobalRotScaleTrans',

docs/en/user_guides/dataset_prepare.md

@@ -122,7 +122,7 @@ Note that:
       ann_file='./data/waymo/kitti_format/waymo_infos_val.pkl',
       waymo_bin_file='./data/waymo/waymo_format/gt.bin',
       data_root='./data/waymo/waymo_format',
-      file_client_args=file_client_args,
+      backend_args=backend_args,
       convert_kitti_format=True,
       idx2metainfo='data/waymo/waymo_format/idx2metainfo.pkl'
       )

docs/zh_cn/notes/faq.md

@@ -10,7 +10,7 @@
 
   | MMDetection3D 版本 |      MMEngine 版本       |        MMCV 版本        |     MMDetection 版本     |
   | ------------------ | :----------------------: | :---------------------: | :----------------------: |
-  | dev-1.x            | mmengine>=0.4.0, \<1.0.0 | mmcv>=2.0.0rc4, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
+  | dev-1.x            | mmengine>=0.6.0, \<1.0.0 | mmcv>=2.0.0rc4, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc3          | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc2          | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |
   | v1.1.0rc1          | mmengine>=0.1.0, \<1.0.0 | mmcv>=2.0.0rc0, \<2.1.0 | mmdet>=3.0.0rc0, \<3.1.0 |

docs/zh_cn/user_guides/backends_support.md

@@ -8,7 +8,7 @@
 
 ```python
 # set file client backends as Ceph
-file_client_args = dict(
+backend_args = dict(
     backend='petrel',
     path_mapping=dict({
         './data/nuscenes/':
@@ -30,15 +30,15 @@ db_sampler = dict(
         coord_type='LIDAR',
         load_dim=4,
         use_dim=4,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     # set file client for data base sampler to load db info file
-    file_client_args=file_client_args)
+    backend_args=backend_args)
 
 train_pipeline = [
     # set file client for loading training data
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, file_client_args=file_client_args),
+    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, backend_args=backend_args),
     # set file client for loading training data annotations
-    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True, file_client_args=file_client_args),
+    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True, backend_args=backend_args),
     dict(type='ObjectSample', db_sampler=db_sampler),
     dict(
         type='ObjectNoise',
@@ -59,7 +59,7 @@ train_pipeline = [
 ]
 test_pipeline = [
     # set file client for loading validation/testing data
-    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, file_client_args=file_client_args),
+    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4, backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),
@@ -87,11 +87,11 @@ data = dict(
     train=dict(
         type='RepeatDataset',
         times=2,
-        dataset=dict(pipeline=train_pipeline, classes=class_names, file_client_args=file_client_args)),
+        dataset=dict(pipeline=train_pipeline, classes=class_names, backend_args=backend_args)),
     # set file client for loading validation info files (.pkl)
-    val=dict(pipeline=test_pipeline, classes=class_names,file_client_args=file_client_args),
+    val=dict(pipeline=test_pipeline, classes=class_names,backend_args=backend_args),
     # set file client for loading testing info files (.pkl)
-    test=dict(pipeline=test_pipeline, classes=class_names, file_client_args=file_client_args))
+    test=dict(pipeline=test_pipeline, classes=class_names, backend_args=backend_args))
 ```
 
 ## 从 Ceph 读取预训练模型

docs/zh_cn/user_guides/config.md

@@ -459,11 +459,11 @@ train_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -484,11 +484,11 @@ test_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug3D',
         img_scale=(1333, 800),

docs/zh_cn/user_guides/data_pipeline.md

@@ -20,11 +20,11 @@ train_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
     dict(
         type='GlobalRotScaleTrans',
@@ -44,11 +44,11 @@ test_pipeline = [
         type='LoadPointsFromFile',
         load_dim=5,
         use_dim=5,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='LoadPointsFromMultiSweeps',
         sweeps_num=10,
-        file_client_args=file_client_args),
+        backend_args=backend_args),
     dict(
         type='MultiScaleFlipAug',
         img_scale=(1333, 800),
@@ -168,11 +168,11 @@ test_pipeline = [
            type='LoadPointsFromFile',
            load_dim=5,
            use_dim=5,
-           file_client_args=file_client_args),
+           backend_args=backend_args),
        dict(
            type='LoadPointsFromMultiSweeps',
            sweeps_num=10,
-           file_client_args=file_client_args),
+           backend_args=backend_args),
        dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
        dict(
            type='GlobalRotScaleTrans',

docs/zh_cn/user_guides/dataset_prepare.md

@@ -121,7 +121,7 @@ python tools/create_data.py waymo --root-path ./data/waymo/ --out-dir ./data/way
       ann_file='./data/waymo/kitti_format/waymo_infos_val.pkl',
       waymo_bin_file='./data/waymo/waymo_format/gt.bin',
       data_root='./data/waymo/waymo_format',
-      file_client_args=file_client_args,
+      backend_args=backend_args,
       convert_kitti_format=True,
       idx2metainfo='data/waymo/waymo_format/idx2metainfo.pkl'
       )

mmdet3d/__init__.py

@@ -6,11 +6,11 @@ from mmengine.utils import digit_version
 
 from .version import __version__, version_info
 
-mmcv_minimum_version = '2.0.0rc0'
+mmcv_minimum_version = '2.0.0rc4'
 mmcv_maximum_version = '2.1.0'
 mmcv_version = digit_version(mmcv.__version__)
 
-mmengine_minimum_version = '0.4.0'
+mmengine_minimum_version = '0.6.0'
 mmengine_maximum_version = '1.0.0'
 mmengine_version = digit_version(mmengine.__version__)
 

mmdet3d/datasets/det3d_dataset.py

@@ -4,7 +4,6 @@ import os
 from os import path as osp
 from typing import Callable, List, Optional, Set, Union
 
-import mmengine
 import numpy as np
 import torch
 from mmengine.dataset import BaseDataset
@@ -61,8 +60,8 @@ class Det3DDataset(BaseDataset):
         load_eval_anns (bool): Whether to load annotations in test_mode,
             the annotation will be save in `eval_ann_infos`, which can be
             used in Evaluator. Defaults to True.
-        file_client_args (dict): Configuration of file client.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
         show_ins_var (bool): For debug purpose. Whether to show variation
             of the number of instances before and after through pipeline.
             Defaults to False.
@@ -80,11 +79,10 @@ class Det3DDataset(BaseDataset):
                  filter_empty_gt: bool = True,
                  test_mode: bool = False,
                  load_eval_anns: bool = True,
-                 file_client_args: dict = dict(backend='disk'),
+                 backend_args: Optional[dict] = None,
                  show_ins_var: bool = False,
                  **kwargs) -> None:
-        # init file client
-        self.file_client = mmengine.FileClient(**file_client_args)
+        self.backend_args = backend_args
         self.filter_empty_gt = filter_empty_gt
         self.load_eval_anns = load_eval_anns
         _default_modality_keys = ('use_lidar', 'use_camera')

mmdet3d/datasets/scannet_dataset.py

@@ -331,7 +331,7 @@ class ScanNetInstanceSegDataset(Seg3DDataset):
                  test_mode: bool = False,
                  ignore_index: Optional[int] = None,
                  scene_idxs: Optional[Union[np.ndarray, str]] = None,
-                 file_client_args: dict = dict(backend='disk'),
+                 backend_args: Optional[dict] = None,
                  **kwargs) -> None:
         super().__init__(
             data_root=data_root,
@@ -343,5 +343,5 @@ class ScanNetInstanceSegDataset(Seg3DDataset):
             test_mode=test_mode,
             ignore_index=ignore_index,
             scene_idxs=scene_idxs,
-            file_client_args=file_client_args,
+            backend_args=backend_args,
             **kwargs)

mmdet3d/datasets/seg3d_dataset.py

@@ -2,9 +2,9 @@
 from os import path as osp
 from typing import Callable, List, Optional, Sequence, Union
 
-import mmengine
 import numpy as np
 from mmengine.dataset import BaseDataset
+from mmengine.fileio import get_local_path
 
 from mmdet3d.registry import DATASETS
 
@@ -49,8 +49,8 @@ class Seg3DDataset(BaseDataset):
         load_eval_anns (bool): Whether to load annotations in test_mode,
             the annotation will be save in `eval_ann_infos`, which can be used
             in Evaluator. Defaults to True.
-        file_client_args (dict): Configuration of file client.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
     """
     METAINFO = {
         'classes': None,  # names of all classes data used for the task
@@ -75,10 +75,9 @@ class Seg3DDataset(BaseDataset):
                  test_mode: bool = False,
                  serialize_data: bool = False,
                  load_eval_anns: bool = True,
-                 file_client_args: dict = dict(backend='disk'),
+                 backend_args: Optional[dict] = None,
                  **kwargs) -> None:
-        # init file client
-        self.file_client = mmengine.FileClient(**file_client_args)
+        self.backend_args = backend_args
         self.modality = modality
         self.load_eval_anns = load_eval_anns
 
@@ -317,7 +316,8 @@ class Seg3DDataset(BaseDataset):
             scene_idxs = np.arange(len(self))
         if isinstance(scene_idxs, str):
             scene_idxs = osp.join(self.data_root, scene_idxs)
-            with self.file_client.get_local_path(scene_idxs) as local_path:
+            with get_local_path(
+                    scene_idxs, backend_args=self.backend_args) as local_path:
                 scene_idxs = np.load(local_path)
         else:
             scene_idxs = np.array(scene_idxs)

mmdet3d/datasets/transforms/dbsampler.py

@@ -5,6 +5,7 @@ from typing import List, Optional
 
 import mmengine
 import numpy as np
+from mmengine.fileio import get_local_path
 
 from mmdet3d.datasets.transforms import data_augment_utils
 from mmdet3d.registry import TRANSFORMS
@@ -91,26 +92,24 @@ class DataBaseSampler(object):
         classes (list[str], optional): List of classes. Defaults to None.
         points_loader (dict): Config of points loader. Defaults to
             dict(type='LoadPointsFromFile', load_dim=4, use_dim=[0, 1, 2, 3]).
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
     """
 
-    def __init__(
-        self,
-        info_path: str,
-        data_root: str,
-        rate: float,
-        prepare: dict,
-        sample_groups: dict,
-        classes: Optional[List[str]] = None,
-        points_loader: dict = dict(
-            type='LoadPointsFromFile',
-            coord_type='LIDAR',
-            load_dim=4,
-            use_dim=[0, 1, 2, 3]),
-        file_client_args: dict = dict(backend='disk')
-    ) -> None:
+    def __init__(self,
+                 info_path: str,
+                 data_root: str,
+                 rate: float,
+                 prepare: dict,
+                 sample_groups: dict,
+                 classes: Optional[List[str]] = None,
+                 points_loader: dict = dict(
+                     type='LoadPointsFromFile',
+                     coord_type='LIDAR',
+                     load_dim=4,
+                     use_dim=[0, 1, 2, 3],
+                     backend_args=None),
+                 backend_args: Optional[dict] = None) -> None:
         super().__init__()
         self.data_root = data_root
         self.info_path = info_path
@@ -120,10 +119,11 @@ class DataBaseSampler(object):
         self.cat2label = {name: i for i, name in enumerate(classes)}
         self.label2cat = {i: name for i, name in enumerate(classes)}
         self.points_loader = TRANSFORMS.build(points_loader)
-        self.file_client = mmengine.FileClient(**file_client_args)
+        self.backend_args = backend_args
 
         # load data base infos
-        with self.file_client.get_local_path(info_path) as local_path:
+        with get_local_path(
+                info_path, backend_args=self.backend_args) as local_path:
             # loading data from a file-like object needs file format
             db_infos = mmengine.load(open(local_path, 'rb'), file_format='pkl')
 

mmdet3d/datasets/transforms/loading.py

@@ -4,11 +4,11 @@ from typing import List, Optional, Union
 
 import mmcv
 import mmengine
-import mmengine.fileio as fileio
 import numpy as np
 from mmcv.transforms import LoadImageFromFile
 from mmcv.transforms.base import BaseTransform
 from mmdet.datasets.transforms import LoadAnnotations
+from mmengine.fileio import get
 
 from mmdet3d.registry import TRANSFORMS
 from mmdet3d.structures.bbox_3d import get_box_type
@@ -25,9 +25,8 @@ class LoadMultiViewImageFromFiles(BaseTransform):
         to_float32 (bool): Whether to convert the img to float32.
             Defaults to False.
         color_type (str): Color type of the file. Defaults to 'unchanged'.
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
         num_views (int): Number of view in a frame. Defaults to 5.
         num_ref_frames (int): Number of frame in loading. Defaults to -1.
         test_mode (bool): Whether is test mode in loading. Defaults to False.
@@ -38,15 +37,14 @@ class LoadMultiViewImageFromFiles(BaseTransform):
     def __init__(self,
                  to_float32: bool = False,
                  color_type: str = 'unchanged',
-                 file_client_args: dict = dict(backend='disk'),
+                 backend_args: Optional[dict] = None,
                  num_views: int = 5,
                  num_ref_frames: int = -1,
                  test_mode: bool = False,
                  set_default_scale: bool = True) -> None:
         self.to_float32 = to_float32
         self.color_type = color_type
-        self.file_client_args = file_client_args.copy()
-        self.file_client = None
+        self.backend_args = backend_args
         self.num_views = num_views
         # num_ref_frames is used for multi-sweep loading
         self.num_ref_frames = num_ref_frames
@@ -164,12 +162,11 @@ class LoadMultiViewImageFromFiles(BaseTransform):
 
         results['ori_cam2img'] = copy.deepcopy(results['cam2img'])
 
-        if self.file_client is None:
-            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
-        img_bytes = [self.file_client.get(name) for name in filename]
+        img_bytes = [
+            get(name, backend_args=self.backend_args) for name in filename
+        ]
         imgs = [
             mmcv.imfrombytes(img_byte, flag=self.color_type)
             for img_byte in img_bytes
@@ -257,13 +254,7 @@ class LoadImageFromFileMono3D(LoadImageFromFile):
                 'nuscenes datasets')
 
         try:
-            if self.file_client_args is not None:
-                file_client = fileio.FileClient.infer_client(
-                    self.file_client_args, filename)
-                img_bytes = file_client.get(filename)
-            else:
-                img_bytes = fileio.get(
-                    filename, backend_args=self.backend_args)
+            img_bytes = get(filename, backend_args=self.backend_args)
             img = mmcv.imfrombytes(
                 img_bytes, flag=self.color_type, backend=self.imdecode_backend)
         except Exception as e:
@@ -331,9 +322,8 @@ class LoadPointsFromMultiSweeps(BaseTransform):
         sweeps_num (int): Number of sweeps. Defaults to 10.
         load_dim (int): Dimension number of the loaded points. Defaults to 5.
         use_dim (list[int]): Which dimension to use. Defaults to [0, 1, 2, 4].
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
         pad_empty_sweeps (bool): Whether to repeat keyframe when
             sweeps is empty. Defaults to False.
         remove_close (bool): Whether to remove close points. Defaults to False.
@@ -345,7 +335,7 @@ class LoadPointsFromMultiSweeps(BaseTransform):
                  sweeps_num: int = 10,
                  load_dim: int = 5,
                  use_dim: List[int] = [0, 1, 2, 4],
-                 file_client_args: dict = dict(backend='disk'),
+                 backend_args: Optional[dict] = None,
                  pad_empty_sweeps: bool = False,
                  remove_close: bool = False,
                  test_mode: bool = False) -> None:
@@ -356,8 +346,7 @@ class LoadPointsFromMultiSweeps(BaseTransform):
         assert max(use_dim) < load_dim, \
             f'Expect all used dimensions < {load_dim}, got {use_dim}'
         self.use_dim = use_dim
-        self.file_client_args = file_client_args.copy()
-        self.file_client = mmengine.FileClient(**self.file_client_args)
+        self.backend_args = backend_args
         self.pad_empty_sweeps = pad_empty_sweeps
         self.remove_close = remove_close
         self.test_mode = test_mode
@@ -371,10 +360,8 @@ class LoadPointsFromMultiSweeps(BaseTransform):
         Returns:
             np.ndarray: An array containing point clouds data.
         """
-        if self.file_client is None:
-            self.file_client = mmengine.FileClient(**self.file_client_args)
         try:
-            pts_bytes = self.file_client.get(pts_filename)
+            pts_bytes = get(pts_filename, backend_args=self.backend_args)
             points = np.frombuffer(pts_bytes, dtype=np.float32)
         except ConnectionError:
             mmengine.check_file_exist(pts_filename)
@@ -592,21 +579,18 @@ class LoadPointsFromFile(BaseTransform):
         use_color (bool): Whether to use color features. Defaults to False.
         norm_intensity (bool): Whether to normlize the intensity. Defaults to
             False.
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
     """
 
-    def __init__(
-        self,
-        coord_type: str,
-        load_dim: int = 6,
-        use_dim: Union[int, List[int]] = [0, 1, 2],
-        shift_height: bool = False,
-        use_color: bool = False,
-        norm_intensity: bool = False,
-        file_client_args: dict = dict(backend='disk')
-    ) -> None:
+    def __init__(self,
+                 coord_type: str,
+                 load_dim: int = 6,
+                 use_dim: Union[int, List[int]] = [0, 1, 2],
+                 shift_height: bool = False,
+                 use_color: bool = False,
+                 norm_intensity: bool = False,
+                 backend_args: Optional[dict] = None) -> None:
         self.shift_height = shift_height
         self.use_color = use_color
         if isinstance(use_dim, int):
@@ -619,8 +603,7 @@ class LoadPointsFromFile(BaseTransform):
         self.load_dim = load_dim
         self.use_dim = use_dim
         self.norm_intensity = norm_intensity
-        self.file_client_args = file_client_args.copy()
-        self.file_client = None
+        self.backend_args = backend_args
 
     def _load_points(self, pts_filename: str) -> np.ndarray:
         """Private function to load point clouds data.
@@ -631,10 +614,8 @@ class LoadPointsFromFile(BaseTransform):
         Returns:
             np.ndarray: An array containing point clouds data.
         """
-        if self.file_client is None:
-            self.file_client = mmengine.FileClient(**self.file_client_args)
         try:
-            pts_bytes = self.file_client.get(pts_filename)
+            pts_bytes = get(pts_filename, backend_args=self.backend_args)
             points = np.frombuffer(pts_bytes, dtype=np.float32)
         except ConnectionError:
             mmengine.check_file_exist(pts_filename)
@@ -698,7 +679,7 @@ class LoadPointsFromFile(BaseTransform):
         repr_str = self.__class__.__name__ + '('
         repr_str += f'shift_height={self.shift_height}, '
         repr_str += f'use_color={self.use_color}, '
-        repr_str += f'file_client_args={self.file_client_args}, '
+        repr_str += f'backend_args={self.backend_args}, '
         repr_str += f'load_dim={self.load_dim}, '
         repr_str += f'use_dim={self.use_dim})'
         return repr_str
@@ -807,37 +788,34 @@ class LoadAnnotations3D(LoadAnnotations):
             panoptic labels. Defaults to None.
         dataset_type (str): Type of dataset used for splitting semantic and
             instance labels. Defaults to None.
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to dict(backend='disk').
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
     """
 
-    def __init__(
-        self,
-        with_bbox_3d: bool = True,
-        with_label_3d: bool = True,
-        with_attr_label: bool = False,
-        with_mask_3d: bool = False,
-        with_seg_3d: bool = False,
-        with_bbox: bool = False,
-        with_label: bool = False,
-        with_mask: bool = False,
-        with_seg: bool = False,
-        with_bbox_depth: bool = False,
-        with_panoptic_3d: bool = False,
-        poly2mask: bool = True,
-        seg_3d_dtype: str = 'np.int64',
-        seg_offset: int = None,
-        dataset_type: str = None,
-        file_client_args: dict = dict(backend='disk')
-    ) -> None:
+    def __init__(self,
+                 with_bbox_3d: bool = True,
+                 with_label_3d: bool = True,
+                 with_attr_label: bool = False,
+                 with_mask_3d: bool = False,
+                 with_seg_3d: bool = False,
+                 with_bbox: bool = False,
+                 with_label: bool = False,
+                 with_mask: bool = False,
+                 with_seg: bool = False,
+                 with_bbox_depth: bool = False,
+                 with_panoptic_3d: bool = False,
+                 poly2mask: bool = True,
+                 seg_3d_dtype: str = 'np.int64',
+                 seg_offset: int = None,
+                 dataset_type: str = None,
+                 backend_args: Optional[dict] = None) -> None:
         super().__init__(
             with_bbox=with_bbox,
             with_label=with_label,
             with_mask=with_mask,
             with_seg=with_seg,
             poly2mask=poly2mask,
-            file_client_args=file_client_args)
+            backend_args=backend_args)
         self.with_bbox_3d = with_bbox_3d
         self.with_bbox_depth = with_bbox_depth
         self.with_label_3d = with_label_3d
@@ -848,7 +826,6 @@ class LoadAnnotations3D(LoadAnnotations):
         self.seg_3d_dtype = eval(seg_3d_dtype)
         self.seg_offset = seg_offset
         self.dataset_type = dataset_type
-        self.file_client = None
 
     def _load_bboxes_3d(self, results: dict) -> dict:
         """Private function to move the 3D bounding box annotation from
@@ -914,10 +891,9 @@ class LoadAnnotations3D(LoadAnnotations):
         """
         pts_instance_mask_path = results['pts_instance_mask_path']
 
-        if self.file_client is None:
-            self.file_client = mmengine.FileClient(**self.file_client_args)
         try:
-            mask_bytes = self.file_client.get(pts_instance_mask_path)
+            mask_bytes = get(
+                pts_instance_mask_path, backend_args=self.backend_args)
             pts_instance_mask = np.frombuffer(mask_bytes, dtype=np.int64)
         except ConnectionError:
             mmengine.check_file_exist(pts_instance_mask_path)
@@ -941,10 +917,9 @@ class LoadAnnotations3D(LoadAnnotations):
         """
         pts_semantic_mask_path = results['pts_semantic_mask_path']
 
-        if self.file_client is None:
-            self.file_client = mmengine.FileClient(**self.file_client_args)
         try:
-            mask_bytes = self.file_client.get(pts_semantic_mask_path)
+            mask_bytes = get(
+                pts_semantic_mask_path, backend_args=self.backend_args)
             # add .copy() to fix read-only bug
             pts_semantic_mask = np.frombuffer(
                 mask_bytes, dtype=self.seg_3d_dtype).copy()
@@ -976,10 +951,9 @@ class LoadAnnotations3D(LoadAnnotations):
         """
         pts_panoptic_mask_path = results['pts_panoptic_mask_path']
 
-        if self.file_client is None:
-            self.file_client = mmengine.FileClient(**self.file_client_args)
         try:
-            mask_bytes = self.file_client.get(pts_panoptic_mask_path)
+            mask_bytes = get(
+                pts_panoptic_mask_path, backend_args=self.backend_args)
             # add .copy() to fix read-only bug
             pts_panoptic_mask = np.frombuffer(
                 mask_bytes, dtype=self.seg_3d_dtype).copy()

mmdet3d/datasets/waymo_dataset.py

@@ -99,7 +99,7 @@ class WaymoDataset(KittiDataset):
         self.cat_ids = range(len(self.METAINFO['classes']))
         self.cat2label = {cat_id: i for i, cat_id in enumerate(self.cat_ids)}
         self.max_sweeps = max_sweeps
-        # we do not provide file_client_args to custom_3d init
+        # we do not provide backend_args to custom_3d init
         # because we want disk loading for info
         # while ceph loading for Prediction2Waymo
         super().__init__(

mmdet3d/engine/hooks/visualization_hook.py

@@ -5,7 +5,7 @@ from typing import Optional, Sequence
 
 import mmcv
 import numpy as np
-from mmengine.fileio import FileClient
+from mmengine.fileio import get
 from mmengine.hooks import Hook
 from mmengine.runner import Runner
 from mmengine.utils import mkdir_or_exist
@@ -44,9 +44,8 @@ class Det3DVisualizationHook(Hook):
         wait_time (float): The interval of show (s). Defaults to 0.
         test_out_dir (str, optional): directory where painted images
             will be saved in testing process.
-        file_client_args (dict): Arguments to instantiate a FileClient.
-            See :class:`mmengine.fileio.FileClient` for details.
-            Defaults to ``dict(backend='disk')``.
+        backend_args (dict, optional): Arguments to instantiate the
+            corresponding backend. Defaults to None.
     """
 
     def __init__(self,
@@ -57,7 +56,7 @@ class Det3DVisualizationHook(Hook):
                  vis_task: str = 'mono_det',
                  wait_time: float = 0.,
                  test_out_dir: Optional[str] = None,
-                 file_client_args: dict = dict(backend='disk')):
+                 backend_args: Optional[dict] = None):
         self._visualizer: Visualizer = Visualizer.get_current_instance()
         self.interval = interval
         self.score_thr = score_thr
@@ -72,8 +71,7 @@ class Det3DVisualizationHook(Hook):
         self.vis_task = vis_task
 
         self.wait_time = wait_time
-        self.file_client_args = file_client_args.copy()
-        self.file_client = None
+        self.backend_args = backend_args
         self.draw = draw
         self.test_out_dir = test_out_dir
         self._test_index = 0
@@ -92,9 +90,6 @@ class Det3DVisualizationHook(Hook):
         if self.draw is False:
             return
 
-        if self.file_client is None:
-            self.file_client = FileClient(**self.file_client_args)
-
         # There is no guarantee that the same batch of images
         # is visualized for each evaluation.
         total_curr_iter = runner.iter + batch_idx
@@ -104,14 +99,14 @@ class Det3DVisualizationHook(Hook):
         # Visualize only the first data
         if 'img_path' in outputs[0]:
             img_path = outputs[0].img_path
-            img_bytes = self.file_client.get(img_path)
+            img_bytes = get(img_path, backend_args=self.backend_args)
             img = mmcv.imfrombytes(img_bytes, channel_order='rgb')
             data_input['img'] = img
 
         if 'lidar_path' in outputs[0]:
             lidar_path = outputs[0].lidar_path
             num_pts_feats = outputs[0].num_pts_feats
-            pts_bytes = self.file_client.get(lidar_path)
+            pts_bytes = get(lidar_path, backend_args=self.backend_args)
             points = np.frombuffer(pts_bytes, dtype=np.float32)
             points = points.reshape(-1, num_pts_feats)
             data_input['points'] = points
@@ -146,23 +141,20 @@ class Det3DVisualizationHook(Hook):
                                          self.test_out_dir)
             mkdir_or_exist(self.test_out_dir)
 
-        if self.file_client is None:
-            self.file_client = FileClient(**self.file_client_args)
-
         for data_sample in outputs:
             self._test_index += 1
 
             data_input = dict()
             if 'img_path' in data_sample:
                 img_path = data_sample.img_path
-                img_bytes = self.file_client.get(img_path)
+                img_bytes = get(img_path, backend_args=self.backend_args)
                 img = mmcv.imfrombytes(img_bytes, channel_order='rgb')
                 data_input['img'] = img
 
             if 'lidar_path' in data_sample:
                 lidar_path = data_sample.lidar_path
                 num_pts_feats = data_sample.num_pts_feats
-                pts_bytes = self.file_client.get(lidar_path)
+                pts_bytes = get(lidar_path, backend_args=self.backend_args)
                 points = np.frombuffer(pts_bytes, dtype=np.float32)
                 points = points.reshape(-1, num_pts_feats)
                 data_input['points'] = points