Merge branch 'add_pipeline_unittest' into 'master'

Add pipeline unittest

See merge request open-mmlab/mmdet.3d!110

mmdet3d/datasets/__init__.py

@@ -6,9 +6,9 @@ from .lyft_dataset import LyftDataset
 from .nuscenes_dataset import NuScenesDataset
 from .pipelines import (GlobalRotScaleTrans, IndoorPointSample,
                         LoadAnnotations3D, LoadPointsFromFile,
-                        NormalizePointsColor, ObjectNoise, ObjectRangeFilter,
-                        ObjectSample, PointShuffle, PointsRangeFilter,
-                        RandomFlip3D)
+                        LoadPointsFromMultiSweeps, NormalizePointsColor,
+                        ObjectNoise, ObjectRangeFilter, ObjectSample,
+                        PointShuffle, PointsRangeFilter, RandomFlip3D)
 from .scannet_dataset import ScanNetDataset
 from .sunrgbd_dataset import SUNRGBDDataset
 
@@ -19,5 +19,6 @@ __all__ = [
     'ObjectSample', 'RandomFlip3D', 'ObjectNoise', 'GlobalRotScaleTrans',
     'PointShuffle', 'ObjectRangeFilter', 'PointsRangeFilter', 'Collect3D',
     'LoadPointsFromFile', 'NormalizePointsColor', 'IndoorPointSample',
-    'LoadAnnotations3D', 'SUNRGBDDataset', 'ScanNetDataset', 'Custom3DDataset'
+    'LoadAnnotations3D', 'SUNRGBDDataset', 'ScanNetDataset', 'Custom3DDataset',
+    'LoadPointsFromMultiSweeps'
 ]

mmdet3d/datasets/pipelines/__init__.py

@@ -2,8 +2,8 @@ from mmdet.datasets.pipelines import Compose
 from .dbsampler import DataBaseSampler
 from .formating import Collect3D, DefaultFormatBundle, DefaultFormatBundle3D
 from .loading import (LoadAnnotations3D, LoadMultiViewImageFromFiles,
-                      LoadPointsFromFile, NormalizePointsColor,
-                      PointSegClassMapping)
+                      LoadPointsFromFile, LoadPointsFromMultiSweeps,
+                      NormalizePointsColor, PointSegClassMapping)
 from .test_time_aug import MultiScaleFlipAug3D
 from .transforms_3d import (GlobalRotScaleTrans, IndoorPointSample,
                             ObjectNoise, ObjectRangeFilter, ObjectSample,
@@ -15,5 +15,5 @@ __all__ = [
     'Compose', 'LoadMultiViewImageFromFiles', 'LoadPointsFromFile',
     'DefaultFormatBundle', 'DefaultFormatBundle3D', 'DataBaseSampler',
     'NormalizePointsColor', 'LoadAnnotations3D', 'IndoorPointSample',
-    'PointSegClassMapping', 'MultiScaleFlipAug3D'
+    'PointSegClassMapping', 'MultiScaleFlipAug3D', 'LoadPointsFromMultiSweeps'
 ]

mmdet3d/datasets/pipelines/loading.py

@@ -230,11 +230,11 @@ class LoadPointsFromFile(object):
         return results
 
     def __repr__(self):
-        repr_str = self.__class__.__name__
-        repr_str += '(shift_height={})'.format(self.shift_height)
-        repr_str += '(mean_color={})'.format(self.color_mean)
-        repr_str += '(load_dim={})'.format(self.load_dim)
-        repr_str += '(use_dim={})'.format(self.use_dim)
+        repr_str = self.__class__.__name__ + '('
+        repr_str += 'shift_height={}, '.format(self.shift_height)
+        repr_str += 'file_client_args={}), '.format(self.file_client_args)
+        repr_str += 'load_dim={}, '.format(self.load_dim)
+        repr_str += 'use_dim={})'.format(self.use_dim)
         return repr_str
 
 
@@ -354,13 +354,13 @@ class LoadAnnotations3D(LoadAnnotations):
     def __repr__(self):
         indent_str = '    '
         repr_str = self.__class__.__name__ + '(\n'
-        repr_str += f'{indent_str}with_bbox_3d={self.with_bbox_3d},\n'
-        repr_str += f'{indent_str}with_label_3d={self.with_label_3d},\n'
-        repr_str += f'{indent_str}with_mask_3d={self.with_mask_3d},\n'
-        repr_str += f'{indent_str}with_seg_3d={self.with_seg_3d},\n'
-        repr_str += f'{indent_str}with_bbox={self.with_bbox},\n'
-        repr_str += f'{indent_str}with_label={self.with_label},\n'
-        repr_str += f'{indent_str}with_mask={self.with_mask},\n'
-        repr_str += f'{indent_str}with_seg={self.with_seg},\n'
+        repr_str += f'{indent_str}with_bbox_3d={self.with_bbox_3d}, '
+        repr_str += f'{indent_str}with_label_3d={self.with_label_3d}, '
+        repr_str += f'{indent_str}with_mask_3d={self.with_mask_3d}, '
+        repr_str += f'{indent_str}with_seg_3d={self.with_seg_3d}, '
+        repr_str += f'{indent_str}with_bbox={self.with_bbox}, '
+        repr_str += f'{indent_str}with_label={self.with_label}, '
+        repr_str += f'{indent_str}with_mask={self.with_mask}, '
+        repr_str += f'{indent_str}with_seg={self.with_seg}, '
         repr_str += f'{indent_str}poly2mask={self.poly2mask})'
         return repr_str

tests/test_pipeline/test_data_augment_utils.py

+import mmcv
+import numpy as np
+
+from mmdet3d.datasets.pipelines.data_augment_utils import (
+    noise_per_object_v3_, points_transform_)
+
+
+def test_noise_per_object_v3_():
+    np.random.seed(0)
+    points = np.fromfile(
+        './tests/data/kitti/training/velodyne_reduced/000000.bin',
+        np.float32).reshape(-1, 4)
+    annos = mmcv.load('./tests/data/kitti/kitti_infos_train.pkl')
+    info = annos[0]
+    annos = info['annos']
+    loc = annos['location']
+    dims = annos['dimensions']
+    rots = annos['rotation_y']
+    gt_bboxes_3d = np.concatenate([loc, dims, rots[..., np.newaxis]],
+                                  axis=1).astype(np.float32)
+
+    noise_per_object_v3_(gt_boxes=gt_bboxes_3d, points=points)
+    expected_gt_bboxes_3d = np.array(
+        [[3.3430212, 2.1475432, 9.388738, 1.2, 1.89, 0.48, 0.05056486]])
+
+    assert points.shape == (800, 4)
+    assert np.allclose(gt_bboxes_3d, expected_gt_bboxes_3d)
+
+
+def test_points_transform():
+    points = np.array([[46.5090, 6.1140, -0.7790, 0.0000],
+                       [42.9490, 6.4050, -0.7050, 0.0000],
+                       [42.9010, 6.5360, -0.7050, 0.0000],
+                       [46.1960, 6.0960, -1.0100, 0.0000],
+                       [43.3080, 6.2680, -0.9360, 0.0000]])
+    gt_boxes = np.array([[
+        1.5340e+01, 8.4691e+00, -1.6855e+00, 1.6400e+00, 3.7000e+00,
+        1.4900e+00, 3.1300e+00
+    ],
+                         [
+                             1.7999e+01, 8.2386e+00, -1.5802e+00, 1.5500e+00,
+                             4.0200e+00, 1.5200e+00, 3.1300e+00
+                         ],
+                         [
+                             2.9620e+01, 8.2617e+00, -1.6185e+00, 1.7800e+00,
+                             4.2500e+00, 1.9000e+00, -3.1200e+00
+                         ],
+                         [
+                             4.8218e+01, 7.8035e+00, -1.3790e+00, 1.6400e+00,
+                             3.7000e+00, 1.5200e+00, -1.0000e-02
+                         ],
+                         [
+                             3.3079e+01, -8.4817e+00, -1.3092e+00, 4.3000e-01,
+                             1.7000e+00, 1.6200e+00, -1.5700e+00
+                         ]])
+    point_masks = np.array([[False, False, False, False, False],
+                            [False, False, False, False, False],
+                            [False, False, False, False, False],
+                            [False, False, False, False, False],
+                            [False, False, False, False, False]])
+    loc_transforms = np.array([[-1.8635, -0.2774, -0.1774],
+                               [-1.0297, -1.0302, -0.3062],
+                               [1.6680, 0.2597, 0.0551],
+                               [0.2230, 0.7257, -0.0097],
+                               [-0.1403, 0.8300, 0.3431]])
+    rot_transforms = np.array([0.6888, -0.3858, 0.1910, -0.0044, -0.0036])
+    valid_mask = np.array([True, True, True, True, True])
+    points_transform_(points, gt_boxes[:, :3], point_masks, loc_transforms,
+                      rot_transforms, valid_mask)
+    assert points.shape == (5, 4)
+    assert gt_boxes.shape == (5, 7)

tests/test_pipeline/test_indoor_sample.py

@@ -52,5 +52,8 @@ def test_indoor_sample():
     sunrgbd_results = sunrgbd_sample_points(sunrgbd_results)
     sunrgbd_choices = np.array([2, 8, 4, 9, 1])
     sunrgbd_points_result = sunrgbd_results['points']
+    repr_str = repr(sunrgbd_sample_points)
+    expected_repr_str = 'IndoorPointSample(num_points=5)'
+    assert repr_str == expected_repr_str
     assert np.allclose(sunrgbd_point_cloud[sunrgbd_choices],
                        sunrgbd_points_result)

tests/test_pipeline/test_loading.py

@@ -4,7 +4,8 @@ import pytest
 from os import path as osp
 
 from mmdet3d.core.bbox import DepthInstance3DBoxes
-from mmdet3d.datasets.pipelines import LoadAnnotations3D, LoadPointsFromFile
+from mmdet3d.datasets.pipelines import (LoadAnnotations3D, LoadPointsFromFile,
+                                        LoadPointsFromMultiSweeps)
 
 
 def test_load_points_from_indoor_file():
@@ -29,6 +30,11 @@ def test_load_points_from_indoor_file():
                                                scannet_info['pts_path'])
     scannet_results = scannet_load_data(scannet_results)
     scannet_point_cloud = scannet_results['points']
+    repr_str = repr(scannet_load_data)
+    expected_repr_str = 'LoadPointsFromFile(shift_height=True, ' \
+                        'file_client_args={\'backend\': \'disk\'}), ' \
+                        'load_dim=6, use_dim=[0, 1, 2])'
+    assert repr_str == expected_repr_str
     assert scannet_point_cloud.shape == (100, 4)
 
 
@@ -93,7 +99,39 @@ def test_load_annotations3D():
 
     scannet_pts_instance_mask = scannet_results['pts_instance_mask']
     scannet_pts_semantic_mask = scannet_results['pts_semantic_mask']
+    repr_str = repr(scannet_load_annotations3D)
+    expected_repr_str = 'LoadAnnotations3D(\n    with_bbox_3d=True,     ' \
+                        'with_label_3d=True,     with_mask_3d=True,     ' \
+                        'with_seg_3d=True,     with_bbox=False,     ' \
+                        'with_label=False,     with_mask=False,     ' \
+                        'with_seg=False,     poly2mask=True)'
+    assert repr_str == expected_repr_str
     assert scannet_gt_boxes.tensor.shape == (27, 7)
     assert scannet_gt_lbaels.shape == (27, )
     assert scannet_pts_instance_mask.shape == (100, )
     assert scannet_pts_semantic_mask.shape == (100, )
+
+
+def test_load_points_from_multi_sweeps():
+    load_points_from_multi_sweeps = LoadPointsFromMultiSweeps()
+    sweep = dict(
+        data_path='./tests/data/nuscenes/sweeps/LIDAR_TOP/'
+        'n008-2018-09-18-12-07-26-0400__LIDAR_TOP__1537287083900561.pcd.bin',
+        timestamp=1537290014899034,
+        sensor2lidar_translation=[-0.02344713, -3.88266051, -0.17151584],
+        sensor2lidar_rotation=np.array(
+            [[9.99979347e-01, 3.99870769e-04, 6.41441690e-03],
+             [-4.42034222e-04, 9.99978299e-01, 6.57316197e-03],
+             [-6.41164929e-03, -6.57586161e-03, 9.99957824e-01]]))
+    results = dict(
+        points=np.array([[1., 2., 3., 4., 5.], [1., 2., 3., 4., 5.],
+                         [1., 2., 3., 4., 5.]]),
+        timestamp=1537290014899034,
+        sweeps=[sweep])
+
+    results = load_points_from_multi_sweeps(results)
+    points = results['points']
+    repr_str = repr(load_points_from_multi_sweeps)
+    expected_repr_str = 'LoadPointsFromMultiSweeps(sweeps_num=10)'
+    assert repr_str == expected_repr_str
+    assert points.shape == (403, 4)

tests/test_pipeline/test_transforms_3d.py

+import mmcv
+import numpy as np
+import torch
+
+from mmdet3d.core import Box3DMode, CameraInstance3DBoxes
+from mmdet3d.datasets import ObjectNoise, ObjectSample
+
+
+def test_remove_points_in_boxes():
+    points = np.array([[68.1370, 3.3580, 2.5160, 0.0000],
+                       [67.6970, 3.5500, 2.5010, 0.0000],
+                       [67.6490, 3.7600, 2.5000, 0.0000],
+                       [66.4140, 3.9010, 2.4590, 0.0000],
+                       [66.0120, 4.0850, 2.4460, 0.0000],
+                       [65.8340, 4.1780, 2.4400, 0.0000],
+                       [65.8410, 4.3860, 2.4400, 0.0000],
+                       [65.7450, 4.5870, 2.4380, 0.0000],
+                       [65.5510, 4.7800, 2.4320, 0.0000],
+                       [65.4860, 4.9820, 2.4300, 0.0000]])
+
+    boxes = np.array(
+        [[30.0285, 10.5110, -1.5304, 0.5100, 0.8700, 1.6000, 1.6400],
+         [7.8369, 1.6053, -1.5605, 0.5800, 1.2300, 1.8200, -3.1000],
+         [10.8740, -1.0827, -1.3310, 0.6000, 0.5200, 1.7100, 1.3500],
+         [14.9783, 2.2466, -1.4950, 0.6100, 0.7300, 1.5300, -1.9200],
+         [11.0656, 0.6195, -1.5202, 0.6600, 1.0100, 1.7600, -1.4600],
+         [10.5994, -7.9049, -1.4980, 0.5300, 1.9600, 1.6800, 1.5600],
+         [28.7068, -8.8244, -1.1485, 0.6500, 1.7900, 1.7500, 3.1200],
+         [20.2630, 5.1947, -1.4799, 0.7300, 1.7600, 1.7300, 1.5100],
+         [18.2496, 3.1887, -1.6109, 0.5600, 1.6800, 1.7100, 1.5600],
+         [7.7396, -4.3245, -1.5801, 0.5600, 1.7900, 1.8000, -0.8300]])
+
+    points = ObjectSample.remove_points_in_boxes(points, boxes)
+    assert points.shape == (10, 4)
+
+
+def test_object_sample():
+    db_sampler = mmcv.ConfigDict({
+        'data_root': './tests/data/kitti/',
+        'info_path': './tests/data/kitti/kitti_dbinfos_train.pkl',
+        'rate': 1.0,
+        'prepare': {
+            'filter_by_difficulty': [-1],
+            'filter_by_min_points': {
+                'Pedestrian': 10
+            }
+        },
+        'classes': ['Pedestrian', 'Cyclist', 'Car'],
+        'sample_groups': {
+            'Pedestrian': 6
+        }
+    })
+    object_sample = ObjectSample(db_sampler)
+    points = np.fromfile(
+        './tests/data/kitti/training/velodyne_reduced/000000.bin',
+        np.float32).reshape(-1, 4)
+    annos = mmcv.load('./tests/data/kitti/kitti_infos_train.pkl')
+    info = annos[0]
+    rect = info['calib']['R0_rect'].astype(np.float32)
+    Trv2c = info['calib']['Tr_velo_to_cam'].astype(np.float32)
+    annos = info['annos']
+    loc = annos['location']
+    dims = annos['dimensions']
+    rots = annos['rotation_y']
+    gt_names = annos['name']
+
+    gt_bboxes_3d = np.concatenate([loc, dims, rots[..., np.newaxis]],
+                                  axis=1).astype(np.float32)
+    gt_bboxes_3d = CameraInstance3DBoxes(gt_bboxes_3d).convert_to(
+        Box3DMode.LIDAR, np.linalg.inv(rect @ Trv2c))
+    CLASSES = ('car', 'pedestrian', 'cyclist')
+    gt_labels = []
+    for cat in gt_names:
+        if cat in CLASSES:
+            gt_labels.append(CLASSES.index(cat))
+        else:
+            gt_labels.append(-1)
+    input_dict = dict(
+        points=points, gt_bboxes_3d=gt_bboxes_3d, gt_labels_3d=gt_labels)
+    input_dict = object_sample(input_dict)
+    points = input_dict['points']
+    gt_bboxes_3d = input_dict['gt_bboxes_3d']
+    gt_labels_3d = input_dict['gt_labels_3d']
+
+    expected_gt_bboxes_3d = torch.tensor(
+        [[8.7314, -1.8559, -1.5997, 0.4800, 1.2000, 1.8900, 0.0100],
+         [8.7314, -1.8559, -1.5997, 0.4800, 1.2000, 1.8900, 0.0100]])
+    expected_gt_labels_3d = np.array([-1, 0])
+    repr_str = repr(object_sample)
+    expected_repr_str = 'ObjectSample'
+    assert repr_str == expected_repr_str
+    assert points.shape == (1177, 4)
+    assert torch.allclose(gt_bboxes_3d.tensor, expected_gt_bboxes_3d)
+    assert np.all(gt_labels_3d == expected_gt_labels_3d)
+
+
+def test_object_noise():
+    np.random.seed(0)
+    object_noise = ObjectNoise()
+    points = np.fromfile(
+        './tests/data/kitti/training/velodyne_reduced/000000.bin',
+        np.float32).reshape(-1, 4)
+    annos = mmcv.load('./tests/data/kitti/kitti_infos_train.pkl')
+    info = annos[0]
+    rect = info['calib']['R0_rect'].astype(np.float32)
+    Trv2c = info['calib']['Tr_velo_to_cam'].astype(np.float32)
+    annos = info['annos']
+    loc = annos['location']
+    dims = annos['dimensions']
+    rots = annos['rotation_y']
+    gt_bboxes_3d = np.concatenate([loc, dims, rots[..., np.newaxis]],
+                                  axis=1).astype(np.float32)
+    gt_bboxes_3d = CameraInstance3DBoxes(gt_bboxes_3d).convert_to(
+        Box3DMode.LIDAR, np.linalg.inv(rect @ Trv2c))
+    input_dict = dict(points=points, gt_bboxes_3d=gt_bboxes_3d)
+    input_dict = object_noise(input_dict)
+    points = input_dict['points']
+    gt_bboxes_3d = input_dict['gt_bboxes_3d'].tensor
+    expected_gt_bboxes_3d = torch.tensor(
+        [[9.1724, -1.7559, -1.3550, 0.4800, 1.2000, 1.8900, 0.0505]])
+    repr_str = repr(object_noise)
+    expected_repr_str = 'ObjectNoise(num_try=100, ' \
+                        'translation_std=[0.25, 0.25, 0.25], ' \
+                        'global_rot_range=[0.0, 0.0], ' \
+                        'rot_range=[-0.15707963267, 0.15707963267])'
+
+    assert repr_str == expected_repr_str
+    assert points.shape == (800, 4)
+    assert torch.allclose(gt_bboxes_3d, expected_gt_bboxes_3d, 1e-3)