Merge branch 'add-tta' into 'master'

Support test time augmentation

See merge request open-mmlab/mmdet.3d!70

mmdet3d/models/necks/second_fpn.py

@@ -100,7 +100,7 @@ class SECONDFusionFPN(SECONDFPN):
                 coors=None,
                 points=None,
                 img_feats=None,
-                img_meta=None):
+                img_metas=None):
         assert len(x) == len(self.in_channels)
         ups = [deblock(x[i]) for i, deblock in enumerate(self.deblocks)]
 
@@ -119,5 +119,5 @@ class SECONDFusionFPN(SECONDFPN):
                 coors[:, 3] / self.downsample_rates[2])
             # fusion for each point
             out = self.fusion_layer(img_feats, points, out,
-                                    downsample_pts_coors, img_meta)
+                                    downsample_pts_coors, img_metas)
         return [out]

mmdet3d/models/roi_heads/base_3droi_head.py

@@ -51,7 +51,7 @@ class Base3DRoIHead(nn.Module, metaclass=ABCMeta):
     @abstractmethod
     def forward_train(self,
                       x,
-                      img_meta,
+                      img_metas,
                       proposal_list,
                       gt_bboxes,
                       gt_labels,
@@ -64,7 +64,7 @@ class Base3DRoIHead(nn.Module, metaclass=ABCMeta):
     def simple_test(self,
                     x,
                     proposal_list,
-                    img_meta,
+                    img_metas,
                     proposals=None,
                     rescale=False,
                     **kwargs):

mmdet3d/models/roi_heads/bbox_heads/parta2_bbox_head.py

@@ -441,7 +441,7 @@ class PartA2BboxHead(nn.Module):
                    bbox_pred,
                    class_labels,
                    class_pred,
-                   img_meta,
+                   img_metas,
                    cfg=None):
         roi_batch_id = rois[..., 0]
         roi_boxes = rois[..., 1:]  # boxes without batch id
@@ -474,8 +474,8 @@ class PartA2BboxHead(nn.Module):
             selected_scores = cur_cls_score[selected]
 
             result_list.append(
-                (img_meta[batch_id]['box_type_3d'](selected_bboxes,
-                                                   self.bbox_coder.code_size),
+                (img_metas[batch_id]['box_type_3d'](selected_bboxes,
+                                                    self.bbox_coder.code_size),
                  selected_scores, selected_label_preds))
         return result_list
 

mmdet3d/models/roi_heads/part_aggregation_roi_head.py

@@ -59,7 +59,7 @@ class PartAggregationROIHead(Base3DRoIHead):
         return hasattr(self,
                        'semantic_head') and self.semantic_head is not None
 
-    def forward_train(self, feats_dict, voxels_dict, img_meta, proposal_list,
+    def forward_train(self, feats_dict, voxels_dict, img_metas, proposal_list,
                       gt_bboxes_3d, gt_labels_3d):
         """Training forward function of PartAggregationROIHead
 
@@ -97,7 +97,7 @@ class PartAggregationROIHead(Base3DRoIHead):
 
         return losses
 
-    def simple_test(self, feats_dict, voxels_dict, img_meta, proposal_list,
+    def simple_test(self, feats_dict, voxels_dict, img_metas, proposal_list,
                     **kwargs):
         """Simple testing forward function of PartAggregationROIHead
 
@@ -131,7 +131,7 @@ class PartAggregationROIHead(Base3DRoIHead):
             bbox_results['bbox_pred'],
             labels_3d,
             cls_preds,
-            img_meta,
+            img_metas,
             cfg=self.test_cfg)
 
         bbox_results = [

mmdet3d/models/voxel_encoders/voxel_encoder.py

@@ -188,7 +188,7 @@ class DynamicVFE(nn.Module):
                 coors,
                 points=None,
                 img_feats=None,
-                img_meta=None):
+                img_metas=None):
         """Forward functions
 
         Args:
@@ -198,7 +198,7 @@ class DynamicVFE(nn.Module):
                 multi-modality fusion. Defaults to None.
             img_feats (list[torch.Tensor], optional): Image fetures used for
                 multi-modality fusion. Defaults to None.
-            img_meta (dict, optional): [description]. Defaults to None.
+            img_metas (dict, optional): [description]. Defaults to None.
 
         Returns:
             tuple: If `return_point_feats` is False, returns voxel features and
@@ -237,7 +237,7 @@ class DynamicVFE(nn.Module):
             if (i == len(self.vfe_layers) - 1 and self.fusion_layer is not None
                     and img_feats is not None):
                 point_feats = self.fusion_layer(img_feats, points, point_feats,
-                                                img_meta)
+                                                img_metas)
             voxel_feats, voxel_coors = self.vfe_scatter(point_feats, coors)
             if i != len(self.vfe_layers) - 1:
                 # need to concat voxel feats if it is not the last vfe
@@ -351,7 +351,7 @@ class HardVFE(nn.Module):
                 num_points,
                 coors,
                 img_feats=None,
-                img_meta=None):
+                img_metas=None):
         """Forward functions
 
         Args:
@@ -360,7 +360,7 @@ class HardVFE(nn.Module):
             coors (torch.Tensor): Coordinates of voxels, shape is Mx(1+NDim).
             img_feats (list[torch.Tensor], optional): Image fetures used for
                 multi-modality fusion. Defaults to None.
-            img_meta (dict, optional): [description]. Defaults to None.
+            img_metas (dict, optional): [description]. Defaults to None.
 
         Returns:
             tuple: If `return_point_feats` is False, returns voxel features and
@@ -410,12 +410,12 @@ class HardVFE(nn.Module):
 
         if (self.fusion_layer is not None and img_feats is not None):
             voxel_feats = self.fusion_with_mask(features, mask, voxel_feats,
-                                                coors, img_feats, img_meta)
+                                                coors, img_feats, img_metas)
 
         return voxel_feats
 
     def fusion_with_mask(self, features, mask, voxel_feats, coors, img_feats,
-                         img_meta):
+                         img_metas):
         """Fuse image and point features with mask.
 
         Args:
@@ -425,7 +425,7 @@ class HardVFE(nn.Module):
             voxel_feats (torch.Tensor): Features of voxels.
             coors (torch.Tensor): Coordinates of each single voxel.
             img_feats (list[torch.Tensor]): Multi-scale feature maps of image.
-            img_meta (list(dict)): Meta information of image and points.
+            img_metas (list(dict)): Meta information of image and points.
 
         Returns:
             torch.Tensor: Fused features of each voxel.
@@ -439,7 +439,7 @@ class HardVFE(nn.Module):
 
         point_feats = voxel_feats[mask]
         point_feats = self.fusion_layer(img_feats, points, point_feats,
-                                        img_meta)
+                                        img_metas)
 
         voxel_canvas = voxel_feats.new_zeros(
             size=(voxel_feats.size(0), voxel_feats.size(1),

tests/test_dataset/test_scannet_dataset.py

@@ -27,7 +27,7 @@ def test_getitem():
         dict(type='IndoorPointSample', num_points=5),
         dict(type='IndoorFlipData', flip_ratio_yz=1.0, flip_ratio_xz=1.0),
         dict(
-            type='IndoorGlobalRotScale',
+            type='IndoorGlobalRotScaleTrans',
             shift_height=True,
             rot_range=[-1 / 36, 1 / 36],
             scale_range=None),
@@ -50,11 +50,11 @@ def test_getitem():
     gt_labels = data['gt_labels_3d']._data
     pts_semantic_mask = data['pts_semantic_mask']._data
     pts_instance_mask = data['pts_instance_mask']._data
-    file_name = data['img_meta']._data['file_name']
-    flip_xz = data['img_meta']._data['flip_xz']
-    flip_yz = data['img_meta']._data['flip_yz']
-    rot_angle = data['img_meta']._data['rot_angle']
-    sample_idx = data['img_meta']._data['sample_idx']
+    file_name = data['img_metas']._data['file_name']
+    flip_xz = data['img_metas']._data['flip_xz']
+    flip_yz = data['img_metas']._data['flip_yz']
+    rot_angle = data['img_metas']._data['rot_angle']
+    sample_idx = data['img_metas']._data['sample_idx']
     assert file_name == './tests/data/scannet/' \
                         'points/scene0000_00.bin'
     assert flip_xz is True

tests/test_dataset/test_sunrgbd_dataset.py

@@ -19,7 +19,7 @@ def test_getitem():
         dict(type='LoadAnnotations3D'),
         dict(type='IndoorFlipData', flip_ratio_yz=1.0),
         dict(
-            type='IndoorGlobalRotScale',
+            type='IndoorGlobalRotScaleTrans',
             shift_height=True,
             rot_range=[-1 / 6, 1 / 6],
             scale_range=[0.85, 1.15]),
@@ -39,12 +39,12 @@ def test_getitem():
     points = data['points']._data
     gt_bboxes_3d = data['gt_bboxes_3d']._data
     gt_labels_3d = data['gt_labels_3d']._data
-    file_name = data['img_meta']._data['file_name']
-    flip_xz = data['img_meta']._data['flip_xz']
-    flip_yz = data['img_meta']._data['flip_yz']
-    scale_ratio = data['img_meta']._data['scale_ratio']
-    rot_angle = data['img_meta']._data['rot_angle']
-    sample_idx = data['img_meta']._data['sample_idx']
+    file_name = data['img_metas']._data['file_name']
+    flip_xz = data['img_metas']._data['flip_xz']
+    flip_yz = data['img_metas']._data['flip_yz']
+    scale_ratio = data['img_metas']._data['scale_ratio']
+    rot_angle = data['img_metas']._data['rot_angle']
+    sample_idx = data['img_metas']._data['sample_idx']
     assert file_name == './tests/data/sunrgbd' \
                         '/points/000001.bin'
     assert flip_xz is False

tests/test_heads.py

@@ -71,7 +71,7 @@ def test_anchor3d_head_loss():
     if not torch.cuda.is_available():
         pytest.skip('test requires GPU and torch+cuda')
     bbox_head_cfg = _get_head_cfg(
-        'second/dv_second_secfpn_2x8_cosine_80e_kitti-3d-3class.py')
+        'second/hv_second_secfpn_6x8_80e_kitti-3d-3class.py')
 
     from mmdet3d.models.builder import build_head
     self = build_head(bbox_head_cfg)
@@ -123,7 +123,7 @@ def test_anchor3d_head_getboxes():
     if not torch.cuda.is_available():
         pytest.skip('test requires GPU and torch+cuda')
     bbox_head_cfg = _get_head_cfg(
-        'second/dv_second_secfpn_2x8_cosine_80e_kitti-3d-3class.py')
+        'second/hv_second_secfpn_6x8_80e_kitti-3d-3class.py')
 
     from mmdet3d.models.builder import build_head
     self = build_head(bbox_head_cfg)

tests/test_pipeline/test_indoor_augment.py

@@ -2,7 +2,8 @@ import numpy as np
 import torch
 
 from mmdet3d.core.bbox import DepthInstance3DBoxes
-from mmdet3d.datasets.pipelines import IndoorFlipData, IndoorGlobalRotScale
+from mmdet3d.datasets.pipelines import (IndoorFlipData,
+                                        IndoorGlobalRotScaleTrans)
 
 
 def test_indoor_flip_data():
@@ -69,7 +70,7 @@ def test_indoor_flip_data():
 
 def test_global_rot_scale():
     np.random.seed(0)
-    sunrgbd_augment = IndoorGlobalRotScale(
+    sunrgbd_augment = IndoorGlobalRotScaleTrans(
         True, rot_range=[-1 / 6, 1 / 6], scale_range=[0.85, 1.15])
     sunrgbd_results = dict()
     sunrgbd_results['points'] = np.array(
@@ -100,7 +101,7 @@ def test_global_rot_scale():
                           expected_sunrgbd_gt_bboxes_3d, 1e-3)
 
     np.random.seed(0)
-    scannet_augment = IndoorGlobalRotScale(
+    scannet_augment = IndoorGlobalRotScaleTrans(
         True, rot_range=[-1 * 1 / 36, 1 / 36], scale_range=None)
     scannet_results = dict()
     scannet_results['points'] = np.array(

tests/test_pipeline/test_indoor_pipeline.py

@@ -30,7 +30,7 @@ def test_scannet_pipeline():
         dict(type='IndoorPointSample', num_points=5),
         dict(type='IndoorFlipData', flip_ratio_yz=1.0, flip_ratio_xz=1.0),
         dict(
-            type='IndoorGlobalRotScale',
+            type='IndoorGlobalRotScaleTrans',
             shift_height=True,
             rot_range=[-1 / 36, 1 / 36],
             scale_range=None),
@@ -113,7 +113,7 @@ def test_sunrgbd_pipeline():
         dict(type='LoadAnnotations3D'),
         dict(type='IndoorFlipData', flip_ratio_yz=1.0),
         dict(
-            type='IndoorGlobalRotScale',
+            type='IndoorGlobalRotScaleTrans',
             shift_height=True,
             rot_range=[-1 / 6, 1 / 6],
             scale_range=[0.85, 1.15]),

tests/test_pipeline/test_outdoor_pipeline.py

@@ -16,14 +16,14 @@ def test_outdoor_aug_pipeline():
         dict(
             type='ObjectNoise',
             num_try=100,
-            loc_noise_std=[1.0, 1.0, 0.5],
+            translation_std=[1.0, 1.0, 0.5],
             global_rot_range=[0.0, 0.0],
-            rot_uniform_noise=[-0.78539816, 0.78539816]),
+            rot_range=[-0.78539816, 0.78539816]),
         dict(type='RandomFlip3D', flip_ratio=0.5),
         dict(
-            type='GlobalRotScale',
-            rot_uniform_noise=[-0.78539816, 0.78539816],
-            scaling_uniform_noise=[0.95, 1.05]),
+            type='GlobalRotScaleTrans',
+            rot_range=[-0.78539816, 0.78539816],
+            scale_ratio_range=[0.95, 1.05]),
         dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
         dict(type='ObjectRangeFilter', point_cloud_range=point_cloud_range),
         dict(type='PointShuffle'),
@@ -98,7 +98,7 @@ def test_outdoor_aug_pipeline():
         pts_filename='tests/data/kitti/a.bin',
         ann_info=dict(gt_bboxes_3d=gt_bboxes_3d, gt_labels_3d=gt_labels_3d),
         bbox3d_fields=[],
-    )
+        img_fields=[])
 
     output = pipeline(results)
 
@@ -133,10 +133,10 @@ def test_outdoor_velocity_aug_pipeline():
         dict(type='LoadPointsFromFile', load_dim=4, use_dim=4),
         dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
         dict(
-            type='GlobalRotScale',
-            rot_uniform_noise=[-0.3925, 0.3925],
-            scaling_uniform_noise=[0.95, 1.05],
-            trans_normal_noise=[0, 0, 0]),
+            type='GlobalRotScaleTrans',
+            rot_range=[-0.3925, 0.3925],
+            scale_ratio_range=[0.95, 1.05],
+            translation_std=[0, 0, 0]),
         dict(type='RandomFlip3D', flip_ratio=0.5),
         dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
         dict(type='ObjectRangeFilter', point_cloud_range=point_cloud_range),
@@ -197,7 +197,7 @@ def test_outdoor_velocity_aug_pipeline():
         pts_filename='tests/data/kitti/a.bin',
         ann_info=dict(gt_bboxes_3d=gt_bboxes_3d, gt_labels_3d=gt_labels_3d),
         bbox3d_fields=[],
-    )
+        img_fields=[])
 
     output = pipeline(results)
 

tools/test.py

@@ -3,6 +3,7 @@ import os
 
 import mmcv
 import torch
+from mmcv import Config, DictAction
 from mmcv.parallel import MMDataParallel, MMDistributedDataParallel
 from mmcv.runner import get_dist_info, init_dist, load_checkpoint
 from tools.fuse_conv_bn import fuse_module
@@ -13,37 +14,6 @@ from mmdet.apis import multi_gpu_test, set_random_seed, single_gpu_test
 from mmdet.core import wrap_fp16_model
 
 
-class MultipleKVAction(argparse.Action):
-    """
-    argparse action to split an argument into KEY=VALUE form
-    on the first = and append to a dictionary. List options should
-    be passed as comma separated values, i.e KEY=V1,V2,V3
-    """
-
-    def _parse_int_float_bool(self, val):
-        try:
-            return int(val)
-        except ValueError:
-            pass
-        try:
-            return float(val)
-        except ValueError:
-            pass
-        if val.lower() in ['true', 'false']:
-            return True if val.lower() == 'true' else False
-        return val
-
-    def __call__(self, parser, namespace, values, option_string=None):
-        options = {}
-        for kv in values:
-            key, val = kv.split('=', maxsplit=1)
-            val = [self._parse_int_float_bool(v) for v in val.split(',')]
-            if len(val) == 1:
-                val = val[0]
-            options[key] = val
-        setattr(namespace, self.dest, options)
-
-
 def parse_args():
     parser = argparse.ArgumentParser(
         description='MMDet test (and eval) a model')
@@ -82,7 +52,7 @@ def parse_args():
         action='store_true',
         help='whether to set deterministic options for CUDNN backend.')
     parser.add_argument(
-        '--options', nargs='+', action=MultipleKVAction, help='custom options')
+        '--options', nargs='+', action=DictAction, help='custom options')
     parser.add_argument(
         '--launcher',
         choices=['none', 'pytorch', 'slurm', 'mpi'],
@@ -109,7 +79,7 @@ def main():
     if args.out is not None and not args.out.endswith(('.pkl', '.pickle')):
         raise ValueError('The output file must be a pkl file.')
 
-    cfg = mmcv.Config.fromfile(args.config)
+    cfg = Config.fromfile(args.config)
     # set cudnn_benchmark
     if cfg.get('cudnn_benchmark', False):
         torch.backends.cudnn.benchmark = True
@@ -167,7 +137,7 @@ def main():
     rank, _ = get_dist_info()
     if rank == 0:
         if args.out:
-            print('\nwriting results to {}'.format(args.out))
+            print(f'\nwriting results to {args.out}')
             mmcv.dump(outputs, args.out)
         kwargs = {} if args.options is None else args.options
         if args.format_only: