File size: 23,141 Bytes
d7a991a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
# Copyright (c) OpenMMLab. All rights reserved.
import copy

import mmcv
import numpy as np
import torch
from mmcv.utils import build_from_cfg

from mmpose.core.camera import CAMERAS
from mmpose.core.post_processing import fliplr_regression
from mmpose.datasets.builder import PIPELINES


@PIPELINES.register_module()
class GetRootCenteredPose:
    """Zero-center the pose around a given root joint. Optionally, the root
    joint can be removed from the original pose and stored as a separate item.

    Note that the root-centered joints may no longer align with some annotation
    information (e.g. flip_pairs, num_joints, inference_channel, etc.) due to
    the removal of the root joint.

    Args:
        item (str): The name of the pose to apply root-centering.
        root_index (int): Root joint index in the pose.
        visible_item (str): The name of the visibility item.
        remove_root (bool): If true, remove the root joint from the pose
        root_name (str): Optional. If not none, it will be used as the key to
            store the root position separated from the original pose.

    Required keys:
        item

    Modified keys:
        item, visible_item, root_name
    """

    def __init__(self,
                 item,
                 root_index,
                 visible_item=None,
                 remove_root=False,
                 root_name=None):
        self.item = item
        self.root_index = root_index
        self.remove_root = remove_root
        self.root_name = root_name
        self.visible_item = visible_item

    def __call__(self, results):
        assert self.item in results
        joints = results[self.item]
        root_idx = self.root_index

        assert joints.ndim >= 2 and joints.shape[-2] > root_idx,\
            f'Got invalid joint shape {joints.shape}'

        root = joints[..., root_idx:root_idx + 1, :]
        joints = joints - root

        results[self.item] = joints
        if self.root_name is not None:
            results[self.root_name] = root

        if self.remove_root:
            results[self.item] = np.delete(
                results[self.item], root_idx, axis=-2)
            if self.visible_item is not None:
                assert self.visible_item in results
                results[self.visible_item] = np.delete(
                    results[self.visible_item], root_idx, axis=-2)
            # Add a flag to avoid latter transforms that rely on the root
            # joint or the original joint index
            results[f'{self.item}_root_removed'] = True

            # Save the root index which is necessary to restore the global pose
            if self.root_name is not None:
                results[f'{self.root_name}_index'] = self.root_index

        return results


@PIPELINES.register_module()
class NormalizeJointCoordinate:
    """Normalize the joint coordinate with given mean and std.

    Args:
        item (str): The name of the pose to normalize.
        mean (array): Mean values of joint coordinates in shape [K, C].
        std (array): Std values of joint coordinates in shape [K, C].
        norm_param_file (str): Optionally load a dict containing `mean` and
            `std` from a file using `mmcv.load`.

    Required keys:
        item

    Modified keys:
        item
    """

    def __init__(self, item, mean=None, std=None, norm_param_file=None):
        self.item = item
        self.norm_param_file = norm_param_file
        if norm_param_file is not None:
            norm_param = mmcv.load(norm_param_file)
            assert 'mean' in norm_param and 'std' in norm_param
            mean = norm_param['mean']
            std = norm_param['std']
        else:
            assert mean is not None
            assert std is not None

        self.mean = np.array(mean, dtype=np.float32)
        self.std = np.array(std, dtype=np.float32)

    def __call__(self, results):
        assert self.item in results
        results[self.item] = (results[self.item] - self.mean) / self.std
        results[f'{self.item}_mean'] = self.mean.copy()
        results[f'{self.item}_std'] = self.std.copy()
        return results


@PIPELINES.register_module()
class ImageCoordinateNormalization:
    """Normalize the 2D joint coordinate with image width and height. Range [0,
    w] is mapped to [-1, 1], while preserving the aspect ratio.

    Args:
        item (str|list[str]): The name of the pose to normalize.
        norm_camera (bool): Whether to normalize camera intrinsics.
            Default: False.
        camera_param (dict|None): The camera parameter dict. See the camera
            class definition for more details. If None is given, the camera
            parameter will be obtained during processing of each data sample
            with the key "camera_param".

    Required keys:
        item

    Modified keys:
        item (, camera_param)
    """

    def __init__(self, item, norm_camera=False, camera_param=None):
        self.item = item
        if isinstance(self.item, str):
            self.item = [self.item]

        self.norm_camera = norm_camera

        if camera_param is None:
            self.static_camera = False
        else:
            self.static_camera = True
            self.camera_param = camera_param

    def __call__(self, results):
        center = np.array(
            [0.5 * results['image_width'], 0.5 * results['image_height']],
            dtype=np.float32)
        scale = np.array(0.5 * results['image_width'], dtype=np.float32)

        for item in self.item:
            results[item] = (results[item] - center) / scale

        if self.norm_camera:
            if self.static_camera:
                camera_param = copy.deepcopy(self.camera_param)
            else:
                assert 'camera_param' in results, \
                    'Camera parameters are missing.'
                camera_param = results['camera_param']
            assert 'f' in camera_param and 'c' in camera_param
            camera_param['f'] = camera_param['f'] / scale
            camera_param['c'] = (camera_param['c'] - center[:, None]) / scale
            if 'camera_param' not in results:
                results['camera_param'] = dict()
            results['camera_param'].update(camera_param)

        return results


@PIPELINES.register_module()
class CollectCameraIntrinsics:
    """Store camera intrinsics in a 1-dim array, including f, c, k, p.

    Args:
        camera_param (dict|None): The camera parameter dict. See the camera
            class definition for more details. If None is given, the camera
            parameter will be obtained during processing of each data sample
            with the key "camera_param".
        need_distortion (bool): Whether need distortion parameters k and p.
            Default: True.

    Required keys:
        camera_param (if camera parameters are not given in initialization)

    Modified keys:
        intrinsics
    """

    def __init__(self, camera_param=None, need_distortion=True):
        if camera_param is None:
            self.static_camera = False
        else:
            self.static_camera = True
            self.camera_param = camera_param
        self.need_distortion = need_distortion

    def __call__(self, results):
        if self.static_camera:
            camera_param = copy.deepcopy(self.camera_param)
        else:
            assert 'camera_param' in results, 'Camera parameters are missing.'
            camera_param = results['camera_param']
        assert 'f' in camera_param and 'c' in camera_param
        intrinsics = np.concatenate(
            [camera_param['f'].reshape(2), camera_param['c'].reshape(2)])
        if self.need_distortion:
            assert 'k' in camera_param and 'p' in camera_param
            intrinsics = np.concatenate([
                intrinsics, camera_param['k'].reshape(3),
                camera_param['p'].reshape(2)
            ])
        results['intrinsics'] = intrinsics

        return results


@PIPELINES.register_module()
class CameraProjection:
    """Apply camera projection to joint coordinates.

    Args:
        item (str): The name of the pose to apply camera projection.
        mode (str): The type of camera projection, supported options are

            - world_to_camera
            - world_to_pixel
            - camera_to_world
            - camera_to_pixel
        output_name (str|None): The name of the projected pose. If None
            (default) is given, the projected pose will be stored in place.
        camera_type (str): The camera class name (should be registered in
            CAMERA).
        camera_param (dict|None): The camera parameter dict. See the camera
            class definition for more details. If None is given, the camera
            parameter will be obtained during processing of each data sample
            with the key "camera_param".

    Required keys:

        - item
        - camera_param (if camera parameters are not given in initialization)

    Modified keys:
        output_name
    """

    def __init__(self,
                 item,
                 mode,
                 output_name=None,
                 camera_type='SimpleCamera',
                 camera_param=None):
        self.item = item
        self.mode = mode
        self.output_name = output_name
        self.camera_type = camera_type
        allowed_mode = {
            'world_to_camera',
            'world_to_pixel',
            'camera_to_world',
            'camera_to_pixel',
        }
        if mode not in allowed_mode:
            raise ValueError(
                f'Got invalid mode: {mode}, allowed modes are {allowed_mode}')

        if camera_param is None:
            self.static_camera = False
        else:
            self.static_camera = True
            self.camera = self._build_camera(camera_param)

    def _build_camera(self, param):
        cfgs = dict(type=self.camera_type, param=param)
        return build_from_cfg(cfgs, CAMERAS)

    def __call__(self, results):
        assert self.item in results
        joints = results[self.item]

        if self.static_camera:
            camera = self.camera
        else:
            assert 'camera_param' in results, 'Camera parameters are missing.'
            camera = self._build_camera(results['camera_param'])

        if self.mode == 'world_to_camera':
            output = camera.world_to_camera(joints)
        elif self.mode == 'world_to_pixel':
            output = camera.world_to_pixel(joints)
        elif self.mode == 'camera_to_world':
            output = camera.camera_to_world(joints)
        elif self.mode == 'camera_to_pixel':
            output = camera.camera_to_pixel(joints)
        else:
            raise NotImplementedError

        output_name = self.output_name
        if output_name is None:
            output_name = self.item

        results[output_name] = output
        return results


@PIPELINES.register_module()
class RelativeJointRandomFlip:
    """Data augmentation with random horizontal joint flip around a root joint.

    Args:
        item (str|list[str]): The name of the pose to flip.
        flip_cfg (dict|list[dict]): Configurations of the fliplr_regression
            function. It should contain the following arguments:

            - ``center_mode``: The mode to set the center location on the \
                x-axis to flip around.
            - ``center_x`` or ``center_index``: Set the x-axis location or \
                the root joint's index to define the flip center.

            Please refer to the docstring of the fliplr_regression function for
            more details.
        visible_item (str|list[str]): The name of the visibility item which
            will be flipped accordingly along with the pose.
        flip_prob (float): Probability of flip.
        flip_camera (bool): Whether to flip horizontal distortion coefficients.
        camera_param (dict|None): The camera parameter dict. See the camera
            class definition for more details. If None is given, the camera
            parameter will be obtained during processing of each data sample
            with the key "camera_param".

    Required keys:
        item

    Modified keys:
        item (, camera_param)
    """

    def __init__(self,
                 item,
                 flip_cfg,
                 visible_item=None,
                 flip_prob=0.5,
                 flip_camera=False,
                 camera_param=None):
        self.item = item
        self.flip_cfg = flip_cfg
        self.vis_item = visible_item
        self.flip_prob = flip_prob
        self.flip_camera = flip_camera
        if camera_param is None:
            self.static_camera = False
        else:
            self.static_camera = True
            self.camera_param = camera_param

        if isinstance(self.item, str):
            self.item = [self.item]
        if isinstance(self.flip_cfg, dict):
            self.flip_cfg = [self.flip_cfg] * len(self.item)
        assert len(self.item) == len(self.flip_cfg)
        if isinstance(self.vis_item, str):
            self.vis_item = [self.vis_item]

    def __call__(self, results):

        if results.get(f'{self.item}_root_removed', False):
            raise RuntimeError('The transform RelativeJointRandomFlip should '
                               f'not be applied to {self.item} whose root '
                               'joint has been removed and joint indices have '
                               'been changed')

        if np.random.rand() <= self.flip_prob:

            flip_pairs = results['ann_info']['flip_pairs']

            # flip joint coordinates
            for i, item in enumerate(self.item):
                assert item in results
                joints = results[item]

                joints_flipped = fliplr_regression(joints, flip_pairs,
                                                   **self.flip_cfg[i])

                results[item] = joints_flipped

            # flip joint visibility
            for vis_item in self.vis_item:
                assert vis_item in results
                visible = results[vis_item]
                visible_flipped = visible.copy()
                for left, right in flip_pairs:
                    visible_flipped[..., left, :] = visible[..., right, :]
                    visible_flipped[..., right, :] = visible[..., left, :]
                results[vis_item] = visible_flipped

            # flip horizontal distortion coefficients
            if self.flip_camera:
                if self.static_camera:
                    camera_param = copy.deepcopy(self.camera_param)
                else:
                    assert 'camera_param' in results, \
                        'Camera parameters are missing.'
                    camera_param = results['camera_param']
                assert 'c' in camera_param
                camera_param['c'][0] *= -1

                if 'p' in camera_param:
                    camera_param['p'][0] *= -1

                if 'camera_param' not in results:
                    results['camera_param'] = dict()
                results['camera_param'].update(camera_param)

        return results


@PIPELINES.register_module()
class PoseSequenceToTensor:
    """Convert pose sequence from numpy array to Tensor.

    The original pose sequence should have a shape of [T,K,C] or [K,C], where
    T is the sequence length, K and C are keypoint number and dimension. The
    converted pose sequence will have a shape of [KxC, T].

    Args:
        item (str): The name of the pose sequence

    Required keys:
        item

    Modified keys:
        item
    """

    def __init__(self, item):
        self.item = item

    def __call__(self, results):
        assert self.item in results
        seq = results[self.item]

        assert isinstance(seq, np.ndarray)
        assert seq.ndim in {2, 3}

        if seq.ndim == 2:
            seq = seq[None, ...]

        T = seq.shape[0]
        seq = seq.transpose(1, 2, 0).reshape(-1, T)
        results[self.item] = torch.from_numpy(seq)

        return results


@PIPELINES.register_module()
class Generate3DHeatmapTarget:
    """Generate the target 3d heatmap.

    Required keys: 'joints_3d', 'joints_3d_visible', 'ann_info'.
    Modified keys: 'target', and 'target_weight'.

    Args:
        sigma: Sigma of heatmap gaussian.
        joint_indices (list): Indices of joints used for heatmap generation.
            If None (default) is given, all joints will be used.
        max_bound (float): The maximal value of heatmap.
    """

    def __init__(self, sigma=2, joint_indices=None, max_bound=1.0):
        self.sigma = sigma
        self.joint_indices = joint_indices
        self.max_bound = max_bound

    def __call__(self, results):
        """Generate the target heatmap."""
        joints_3d = results['joints_3d']
        joints_3d_visible = results['joints_3d_visible']
        cfg = results['ann_info']
        image_size = cfg['image_size']
        W, H, D = cfg['heatmap_size']
        heatmap3d_depth_bound = cfg['heatmap3d_depth_bound']
        joint_weights = cfg['joint_weights']
        use_different_joint_weights = cfg['use_different_joint_weights']

        # select the joints used for target generation
        if self.joint_indices is not None:
            joints_3d = joints_3d[self.joint_indices, ...]
            joints_3d_visible = joints_3d_visible[self.joint_indices, ...]
            joint_weights = joint_weights[self.joint_indices, ...]
        num_joints = joints_3d.shape[0]

        # get the joint location in heatmap coordinates
        mu_x = joints_3d[:, 0] * W / image_size[0]
        mu_y = joints_3d[:, 1] * H / image_size[1]
        mu_z = (joints_3d[:, 2] / heatmap3d_depth_bound + 0.5) * D

        target = np.zeros([num_joints, D, H, W], dtype=np.float32)

        target_weight = joints_3d_visible[:, 0].astype(np.float32)
        target_weight = target_weight * (mu_z >= 0) * (mu_z < D)
        if use_different_joint_weights:
            target_weight = target_weight * joint_weights
        target_weight = target_weight[:, None]

        # only compute the voxel value near the joints location
        tmp_size = 3 * self.sigma

        # get neighboring voxels coordinates
        x = y = z = np.arange(2 * tmp_size + 1, dtype=np.float32) - tmp_size
        zz, yy, xx = np.meshgrid(z, y, x)
        xx = xx[None, ...].astype(np.float32)
        yy = yy[None, ...].astype(np.float32)
        zz = zz[None, ...].astype(np.float32)
        mu_x = mu_x[..., None, None, None]
        mu_y = mu_y[..., None, None, None]
        mu_z = mu_z[..., None, None, None]
        xx, yy, zz = xx + mu_x, yy + mu_y, zz + mu_z

        # round the coordinates
        xx = xx.round().clip(0, W - 1)
        yy = yy.round().clip(0, H - 1)
        zz = zz.round().clip(0, D - 1)

        # compute the target value near joints
        local_target = \
            np.exp(-((xx - mu_x)**2 + (yy - mu_y)**2 + (zz - mu_z)**2) /
                   (2 * self.sigma**2))

        # put the local target value to the full target heatmap
        local_size = xx.shape[1]
        idx_joints = np.tile(
            np.arange(num_joints)[:, None, None, None],
            [1, local_size, local_size, local_size])
        idx = np.stack([idx_joints, zz, yy, xx],
                       axis=-1).astype(int).reshape(-1, 4)
        target[idx[:, 0], idx[:, 1], idx[:, 2],
               idx[:, 3]] = local_target.reshape(-1)
        target = target * self.max_bound
        results['target'] = target
        results['target_weight'] = target_weight
        return results


@PIPELINES.register_module()
class GenerateVoxel3DHeatmapTarget:
    """Generate the target 3d heatmap.

    Required keys: 'joints_3d', 'joints_3d_visible', 'ann_info_3d'.
    Modified keys: 'target', and 'target_weight'.

    Args:
        sigma: Sigma of heatmap gaussian (mm).
        joint_indices (list): Indices of joints used for heatmap generation.
            If None (default) is given, all joints will be used.
    """

    def __init__(self, sigma=200.0, joint_indices=None):
        self.sigma = sigma  # mm
        self.joint_indices = joint_indices

    def __call__(self, results):
        """Generate the target heatmap."""
        joints_3d = results['joints_3d']
        joints_3d_visible = results['joints_3d_visible']
        cfg = results['ann_info']

        num_people = len(joints_3d)
        num_joints = joints_3d[0].shape[0]

        if self.joint_indices is not None:
            num_joints = len(self.joint_indices)
            joint_indices = self.joint_indices
        else:
            joint_indices = list(range(num_joints))

        space_size = cfg['space_size']
        space_center = cfg['space_center']
        cube_size = cfg['cube_size']
        grids_x = np.linspace(-space_size[0] / 2, space_size[0] / 2,
                              cube_size[0]) + space_center[0]
        grids_y = np.linspace(-space_size[1] / 2, space_size[1] / 2,
                              cube_size[1]) + space_center[1]
        grids_z = np.linspace(-space_size[2] / 2, space_size[2] / 2,
                              cube_size[2]) + space_center[2]

        target = np.zeros(
            (num_joints, cube_size[0], cube_size[1], cube_size[2]),
            dtype=np.float32)

        for n in range(num_people):
            for idx, joint_id in enumerate(joint_indices):
                mu_x = joints_3d[n][joint_id][0]
                mu_y = joints_3d[n][joint_id][1]
                mu_z = joints_3d[n][joint_id][2]
                vis = joints_3d_visible[n][joint_id][0]
                if vis < 1:
                    continue
                i_x = [
                    np.searchsorted(grids_x, mu_x - 3 * self.sigma),
                    np.searchsorted(grids_x, mu_x + 3 * self.sigma, 'right')
                ]
                i_y = [
                    np.searchsorted(grids_y, mu_y - 3 * self.sigma),
                    np.searchsorted(grids_y, mu_y + 3 * self.sigma, 'right')
                ]
                i_z = [
                    np.searchsorted(grids_z, mu_z - 3 * self.sigma),
                    np.searchsorted(grids_z, mu_z + 3 * self.sigma, 'right')
                ]
                if i_x[0] >= i_x[1] or i_y[0] >= i_y[1] or i_z[0] >= i_z[1]:
                    continue
                kernel_xs, kernel_ys, kernel_zs = np.meshgrid(
                    grids_x[i_x[0]:i_x[1]],
                    grids_y[i_y[0]:i_y[1]],
                    grids_z[i_z[0]:i_z[1]],
                    indexing='ij')
                g = np.exp(-((kernel_xs - mu_x)**2 + (kernel_ys - mu_y)**2 +
                             (kernel_zs - mu_z)**2) / (2 * self.sigma**2))
                target[idx, i_x[0]:i_x[1], i_y[0]:i_y[1], i_z[0]:i_z[1]] \
                    = np.maximum(target[idx, i_x[0]:i_x[1],
                                 i_y[0]:i_y[1], i_z[0]:i_z[1]], g)

        target = np.clip(target, 0, 1)
        if target.shape[0] == 1:
            target = target[0]

        results['targets_3d'] = target

        return results