kitti_dataset.py 20.4 KB
Newer Older
zhangwenwei's avatar
zhangwenwei committed
1
2
import copy
import os
3
4
import os.path as osp
import tempfile
zhangwenwei's avatar
zhangwenwei committed
5
6
7
8

import mmcv
import numpy as np
import torch
zhangwenwei's avatar
zhangwenwei committed
9
from mmcv.utils import print_log
zhangwenwei's avatar
zhangwenwei committed
10

zhangwenwei's avatar
zhangwenwei committed
11
from mmdet.datasets import DATASETS
zhangwenwei's avatar
zhangwenwei committed
12
from ..core.bbox import Box3DMode, CameraInstance3DBoxes, points_cam2img
zhangwenwei's avatar
zhangwenwei committed
13
from .custom_3d import Custom3DDataset
zhangwenwei's avatar
zhangwenwei committed
14
15


16
@DATASETS.register_module()
zhangwenwei's avatar
zhangwenwei committed
17
class KittiDataset(Custom3DDataset):
zhangwenwei's avatar
zhangwenwei committed
18
19
20
21

    CLASSES = ('car', 'pedestrian', 'cyclist')

    def __init__(self,
zhangwenwei's avatar
zhangwenwei committed
22
                 data_root,
zhangwenwei's avatar
zhangwenwei committed
23
24
                 ann_file,
                 split,
zhangwenwei's avatar
zhangwenwei committed
25
                 pts_prefix='velodyne',
zhangwenwei's avatar
zhangwenwei committed
26
                 pipeline=None,
zhangwenwei's avatar
zhangwenwei committed
27
                 classes=None,
zhangwenwei's avatar
zhangwenwei committed
28
                 modality=None,
29
30
                 box_type_3d='LiDAR',
                 filter_empty_gt=True,
zhangwenwei's avatar
zhangwenwei committed
31
                 test_mode=False):
zhangwenwei's avatar
zhangwenwei committed
32
33
34
35
36
37
        super().__init__(
            data_root=data_root,
            ann_file=ann_file,
            pipeline=pipeline,
            classes=classes,
            modality=modality,
38
39
            box_type_3d=box_type_3d,
            filter_empty_gt=filter_empty_gt,
zhangwenwei's avatar
zhangwenwei committed
40
41
42
            test_mode=test_mode)

        self.root_split = os.path.join(self.data_root, split)
zhangwenwei's avatar
zhangwenwei committed
43
44
        assert self.modality is not None
        self.pcd_limit_range = [0, -40, -3, 70.4, 40, 0.0]
zhangwenwei's avatar
zhangwenwei committed
45
        self.pts_prefix = pts_prefix
zhangwenwei's avatar
zhangwenwei committed
46

zhangwenwei's avatar
zhangwenwei committed
47
48
49
50
    def _get_pts_filename(self, idx):
        pts_filename = osp.join(self.root_split, self.pts_prefix,
                                f'{idx:06d}.bin')
        return pts_filename
zhangwenwei's avatar
zhangwenwei committed
51

zhangwenwei's avatar
zhangwenwei committed
52
53
    def get_data_info(self, index):
        info = self.data_infos[index]
zhangwenwei's avatar
zhangwenwei committed
54
        sample_idx = info['image']['image_idx']
zhangwenwei's avatar
zhangwenwei committed
55
        img_filename = os.path.join(self.data_root,
zhangwenwei's avatar
zhangwenwei committed
56
57
                                    info['image']['image_path'])

zhangwenwei's avatar
zhangwenwei committed
58
59
60
61
62
63
        # TODO: consider use torch.Tensor only
        rect = info['calib']['R0_rect'].astype(np.float32)
        Trv2c = info['calib']['Tr_velo_to_cam'].astype(np.float32)
        P2 = info['calib']['P2'].astype(np.float32)
        lidar2img = P2 @ rect @ Trv2c

zhangwenwei's avatar
zhangwenwei committed
64
        pts_filename = self._get_pts_filename(sample_idx)
zhangwenwei's avatar
zhangwenwei committed
65
66
        input_dict = dict(
            sample_idx=sample_idx,
zhangwenwei's avatar
zhangwenwei committed
67
            pts_filename=pts_filename,
zhangwenwei's avatar
zhangwenwei committed
68
69
            img_prefix=None,
            img_info=dict(filename=img_filename),
zhangwenwei's avatar
zhangwenwei committed
70
71
72
            lidar2img=lidar2img)

        if not self.test_mode:
zhangwenwei's avatar
zhangwenwei committed
73
            annos = self.get_ann_info(index)
zhangwenwei's avatar
zhangwenwei committed
74
            input_dict['ann_info'] = annos
zhangwenwei's avatar
zhangwenwei committed
75
76
77
78
79

        return input_dict

    def get_ann_info(self, index):
        # Use index to get the annos, thus the evalhook could also use this api
zhangwenwei's avatar
zhangwenwei committed
80
        info = self.data_infos[index]
zhangwenwei's avatar
zhangwenwei committed
81
82
83
84
85
        rect = info['calib']['R0_rect'].astype(np.float32)
        Trv2c = info['calib']['Tr_velo_to_cam'].astype(np.float32)

        annos = info['annos']
        # we need other objects to avoid collision when sample
86
        annos = self.remove_dontcare(annos)
zhangwenwei's avatar
zhangwenwei committed
87
88
89
90
91
92
93
        loc = annos['location']
        dims = annos['dimensions']
        rots = annos['rotation_y']
        gt_names = annos['name']
        # print(gt_names, len(loc))
        gt_bboxes_3d = np.concatenate([loc, dims, rots[..., np.newaxis]],
                                      axis=1).astype(np.float32)
94
95
96

        # convert gt_bboxes_3d to velodyne coordinates
        gt_bboxes_3d = CameraInstance3DBoxes(gt_bboxes_3d).convert_to(
97
            self.box_mode_3d, np.linalg.inv(rect @ Trv2c))
zhangwenwei's avatar
zhangwenwei committed
98
99
100
        gt_bboxes = annos['bbox']

        selected = self.drop_arrays_by_name(gt_names, ['DontCare'])
101
        # gt_bboxes_3d = gt_bboxes_3d[selected].astype('float32')
zhangwenwei's avatar
zhangwenwei committed
102
103
104
105
106
107
108
109
110
111
112
        gt_bboxes = gt_bboxes[selected].astype('float32')
        gt_names = gt_names[selected]

        gt_labels = []
        for cat in gt_names:
            if cat in self.CLASSES:
                gt_labels.append(self.CLASSES.index(cat))
            else:
                gt_labels.append(-1)
        gt_labels = np.array(gt_labels)
        gt_labels_3d = copy.deepcopy(gt_labels)
zhangwenwei's avatar
zhangwenwei committed
113
114
115

        anns_results = dict(
            gt_bboxes_3d=gt_bboxes_3d,
zhangwenwei's avatar
zhangwenwei committed
116
            gt_labels_3d=gt_labels_3d,
zhangwenwei's avatar
zhangwenwei committed
117
            bboxes=gt_bboxes,
liyinhao's avatar
liyinhao committed
118
119
            labels=gt_labels,
            gt_names=gt_names)
zhangwenwei's avatar
zhangwenwei committed
120
121
122
123
124
125
126
127
128
129
130
131
        return anns_results

    def drop_arrays_by_name(self, gt_names, used_classes):
        inds = [i for i, x in enumerate(gt_names) if x not in used_classes]
        inds = np.array(inds, dtype=np.int64)
        return inds

    def keep_arrays_by_name(self, gt_names, used_classes):
        inds = [i for i, x in enumerate(gt_names) if x in used_classes]
        inds = np.array(inds, dtype=np.int64)
        return inds

132
133
134
135
136
137
138
139
140
141
    def remove_dontcare(self, ann_info):
        img_filtered_annotations = {}
        relevant_annotation_indices = [
            i for i, x in enumerate(ann_info['name']) if x != 'DontCare'
        ]
        for key in ann_info.keys():
            img_filtered_annotations[key] = (
                ann_info[key][relevant_annotation_indices])
        return img_filtered_annotations

142
143
144
145
146
147
148
149
150
151
    def format_results(self,
                       outputs,
                       pklfile_prefix=None,
                       submission_prefix=None):
        if pklfile_prefix is None:
            tmp_dir = tempfile.TemporaryDirectory()
            pklfile_prefix = osp.join(tmp_dir.name, 'results')
        else:
            tmp_dir = None

zhangwenwei's avatar
zhangwenwei committed
152
        if not isinstance(outputs[0], dict):
zhangwenwei's avatar
zhangwenwei committed
153
            result_files = self.bbox2result_kitti2d(outputs, self.CLASSES,
zhangwenwei's avatar
zhangwenwei committed
154
                                                    pklfile_prefix,
155
                                                    submission_prefix)
zhangwenwei's avatar
zhangwenwei committed
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
        elif 'pts_bbox' in outputs[0] or 'img_bbox' in outputs[0]:
            result_files = dict()
            for name in outputs[0]:
                results_ = [out[name] for out in outputs]
                pklfile_prefix_ = pklfile_prefix + name
                if submission_prefix is not None:
                    submission_prefix_ = submission_prefix + name
                else:
                    submission_prefix_ = None
                if 'img' in name:
                    result_files = self.bbox2result_kitti2d(
                        results_, self.CLASSES, pklfile_prefix_,
                        submission_prefix_)
                else:
                    result_files_ = self.bbox2result_kitti(
                        results_, self.CLASSES, pklfile_prefix_,
                        submission_prefix_)
                result_files[name] = result_files_
zhangwenwei's avatar
zhangwenwei committed
174
        else:
zhangwenwei's avatar
zhangwenwei committed
175
            result_files = self.bbox2result_kitti(outputs, self.CLASSES,
176
177
                                                  pklfile_prefix,
                                                  submission_prefix)
zhangwenwei's avatar
zhangwenwei committed
178
        return result_files, tmp_dir
zhangwenwei's avatar
zhangwenwei committed
179

180
181
182
183
184
    def evaluate(self,
                 results,
                 metric=None,
                 logger=None,
                 pklfile_prefix=None,
zhangwenwei's avatar
zhangwenwei committed
185
                 submission_prefix=None):
186
187
188
189
190
191
192
193
194
195
196
197
198
199
        """Evaluation in KITTI protocol.

        Args:
            results (list): Testing results of the dataset.
            metric (str | list[str]): Metrics to be evaluated.
            logger (logging.Logger | str | None): Logger used for printing
                related information during evaluation. Default: None.
            pklfile_prefix (str | None): The prefix of pkl files. It includes
                the file path and the prefix of filename, e.g., "a/b/prefix".
                If not specified, a temp file will be created. Default: None.
            submission_prefix (str | None): The prefix of submission datas.
                If not specified, the submission data will not be generated.

        Returns:
zhangwenwei's avatar
zhangwenwei committed
200
            dict[str: float]: results of each evaluation metric
201
202
        """
        result_files, tmp_dir = self.format_results(results, pklfile_prefix)
zhangwenwei's avatar
zhangwenwei committed
203
        from mmdet3d.core.evaluation import kitti_eval
zhangwenwei's avatar
zhangwenwei committed
204
        gt_annos = [info['annos'] for info in self.data_infos]
zhangwenwei's avatar
zhangwenwei committed
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222

        if isinstance(result_files, dict):
            ap_dict = dict()
            for name, result_files_ in result_files.items():
                eval_types = ['bbox', 'bev', '3d']
                if 'img' in name:
                    eval_types = ['bbox']
                ap_result_str, ap_dict_ = kitti_eval(
                    gt_annos,
                    result_files_,
                    self.CLASSES,
                    eval_types=eval_types)
                for ap_type, ap in ap_dict_.items():
                    ap_dict[f'{name}/{ap_type}'] = float('{:.4f}'.format(ap))

                print_log(
                    f'Results of {name}:\n' + ap_result_str, logger=logger)

zhangwenwei's avatar
zhangwenwei committed
223
        else:
zhangwenwei's avatar
zhangwenwei committed
224
225
226
227
228
229
230
231
            if metric == 'img_bbox':
                ap_result_str, ap_dict = kitti_eval(
                    gt_annos, result_files, self.CLASSES, eval_types=['bbox'])
            else:
                ap_result_str, ap_dict = kitti_eval(gt_annos, result_files,
                                                    self.CLASSES)
            print_log('\n' + ap_result_str, logger=logger)

232
233
        if tmp_dir is not None:
            tmp_dir.cleanup()
234
        return ap_dict
235
236
237
238
239
240

    def bbox2result_kitti(self,
                          net_outputs,
                          class_names,
                          pklfile_prefix=None,
                          submission_prefix=None):
zhangwenwei's avatar
zhangwenwei committed
241
        assert len(net_outputs) == len(self.data_infos)
242
243
        if submission_prefix is not None:
            mmcv.mkdir_or_exist(submission_prefix)
zhangwenwei's avatar
zhangwenwei committed
244
245

        det_annos = []
zhangwenwei's avatar
zhangwenwei committed
246
        print('\nConverting prediction to KITTI format')
zhangwenwei's avatar
zhangwenwei committed
247
248
249
        for idx, pred_dicts in enumerate(
                mmcv.track_iter_progress(net_outputs)):
            annos = []
zhangwenwei's avatar
zhangwenwei committed
250
            info = self.data_infos[idx]
zhangwenwei's avatar
zhangwenwei committed
251
            sample_idx = info['image']['image_idx']
zhangwenwei's avatar
zhangwenwei committed
252
            image_shape = info['image']['image_shape'][:2]
zhangwenwei's avatar
zhangwenwei committed
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

            box_dict = self.convert_valid_bboxes(pred_dicts, info)
            if len(box_dict['bbox']) > 0:
                box_2d_preds = box_dict['bbox']
                box_preds = box_dict['box3d_camera']
                scores = box_dict['scores']
                box_preds_lidar = box_dict['box3d_lidar']
                label_preds = box_dict['label_preds']

                anno = {
                    'name': [],
                    'truncated': [],
                    'occluded': [],
                    'alpha': [],
                    'bbox': [],
                    'dimensions': [],
                    'location': [],
                    'rotation_y': [],
                    'score': []
                }

                for box, box_lidar, bbox, score, label in zip(
                        box_preds, box_preds_lidar, box_2d_preds, scores,
                        label_preds):
                    bbox[2:] = np.minimum(bbox[2:], image_shape[::-1])
                    bbox[:2] = np.maximum(bbox[:2], [0, 0])
                    anno['name'].append(class_names[int(label)])
                    anno['truncated'].append(0.0)
                    anno['occluded'].append(0)
                    anno['alpha'].append(
                        -np.arctan2(-box_lidar[1], box_lidar[0]) + box[6])
                    anno['bbox'].append(bbox)
                    anno['dimensions'].append(box[3:6])
                    anno['location'].append(box[:3])
                    anno['rotation_y'].append(box[6])
                    anno['score'].append(score)

                anno = {k: np.stack(v) for k, v in anno.items()}
                annos.append(anno)

                if submission_prefix is not None:
                    curr_file = f'{submission_prefix}/{sample_idx:06d}.txt'
                    with open(curr_file, 'w') as f:
                        bbox = anno['bbox']
                        loc = anno['location']
                        dims = anno['dimensions']  # lhw -> hwl

                        for idx in range(len(bbox)):
                            print(
                                '{} -1 -1 {:.4f} {:.4f} {:.4f} {:.4f} '
                                '{:.4f} {:.4f} {:.4f} '
                                '{:.4f} {:.4f} {:.4f} {:.4f} {:.4f} {:.4f}'.
                                format(anno['name'][idx], anno['alpha'][idx],
                                       bbox[idx][0], bbox[idx][1],
                                       bbox[idx][2], bbox[idx][3],
                                       dims[idx][1], dims[idx][2],
                                       dims[idx][0], loc[idx][0], loc[idx][1],
                                       loc[idx][2], anno['rotation_y'][idx],
                                       anno['score'][idx]),
                                file=f)
            else:
                annos.append({
                    'name': np.array([]),
                    'truncated': np.array([]),
                    'occluded': np.array([]),
                    'alpha': np.array([]),
                    'bbox': np.zeros([0, 4]),
                    'dimensions': np.zeros([0, 3]),
                    'location': np.zeros([0, 3]),
                    'rotation_y': np.array([]),
                    'score': np.array([]),
                })
            annos[-1]['sample_idx'] = np.array(
                [sample_idx] * len(annos[-1]['score']), dtype=np.int64)
zhangwenwei's avatar
zhangwenwei committed
327
328
329

            det_annos += annos

330
331
332
        if pklfile_prefix is not None:
            if not pklfile_prefix.endswith(('.pkl', '.pickle')):
                out = f'{pklfile_prefix}.pkl'
zhangwenwei's avatar
zhangwenwei committed
333
334
335
336
337
338
339
340
            mmcv.dump(det_annos, out)
            print('Result is saved to %s' % out)

        return det_annos

    def bbox2result_kitti2d(self,
                            net_outputs,
                            class_names,
341
342
                            pklfile_prefix=None,
                            submission_prefix=None):
liyinhao's avatar
liyinhao committed
343
        """Convert results to kitti format for evaluation and test submission.
zhangwenwei's avatar
zhangwenwei committed
344
345
346
347

        Args:
            net_outputs (List[array]): list of array storing the bbox and score
            class_nanes (List[String]): A list of class names
348
349
            pklfile_prefix (str | None): The prefix of pkl file.
            submission_prefix (str | None): The prefix of submission file.
zhangwenwei's avatar
zhangwenwei committed
350
351

        Return:
liyinhao's avatar
liyinhao committed
352
            List[dict]: A list of dict have the kitti format
zhangwenwei's avatar
zhangwenwei committed
353
        """
zhangwenwei's avatar
zhangwenwei committed
354
        assert len(net_outputs) == len(self.data_infos)
zhangwenwei's avatar
zhangwenwei committed
355
356

        det_annos = []
zhangwenwei's avatar
zhangwenwei committed
357
        print('\nConverting prediction to KITTI format')
zhangwenwei's avatar
zhangwenwei committed
358
359
360
361
362
363
364
365
366
367
368
369
370
        for i, bboxes_per_sample in enumerate(
                mmcv.track_iter_progress(net_outputs)):
            annos = []
            anno = dict(
                name=[],
                truncated=[],
                occluded=[],
                alpha=[],
                bbox=[],
                dimensions=[],
                location=[],
                rotation_y=[],
                score=[])
zhangwenwei's avatar
zhangwenwei committed
371
            sample_idx = self.data_infos[i]['image']['image_idx']
zhangwenwei's avatar
zhangwenwei committed
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

            num_example = 0
            for label in range(len(bboxes_per_sample)):
                bbox = bboxes_per_sample[label]
                for i in range(bbox.shape[0]):
                    anno['name'].append(class_names[int(label)])
                    anno['truncated'].append(0.0)
                    anno['occluded'].append(0)
                    anno['alpha'].append(0.0)
                    anno['bbox'].append(bbox[i, :4])
                    # set dimensions (height, width, length) to zero
                    anno['dimensions'].append(
                        np.zeros(shape=[3], dtype=np.float32))
                    # set the 3D translation to (-1000, -1000, -1000)
                    anno['location'].append(
                        np.ones(shape=[3], dtype=np.float32) * (-1000.0))
                    anno['rotation_y'].append(0.0)
                    anno['score'].append(bbox[i, 4])
                    num_example += 1

            if num_example == 0:
                annos.append(
                    dict(
                        name=np.array([]),
                        truncated=np.array([]),
                        occluded=np.array([]),
                        alpha=np.array([]),
                        bbox=np.zeros([0, 4]),
                        dimensions=np.zeros([0, 3]),
                        location=np.zeros([0, 3]),
                        rotation_y=np.array([]),
                        score=np.array([]),
                    ))
            else:
                anno = {k: np.stack(v) for k, v in anno.items()}
                annos.append(anno)

            annos[-1]['sample_idx'] = np.array(
                [sample_idx] * num_example, dtype=np.int64)
            det_annos += annos

413
414
415
416
417
418
419
420
        if pklfile_prefix is not None:
            # save file in pkl format
            pklfile_path = (
                pklfile_prefix[:-4] if pklfile_prefix.endswith(
                    ('.pkl', '.pickle')) else pklfile_prefix)
            mmcv.dump(det_annos, pklfile_path)

        if submission_prefix is not None:
zhangwenwei's avatar
zhangwenwei committed
421
            # save file in submission format
422
423
            mmcv.mkdir_or_exist(submission_prefix)
            print(f'Saving KITTI submission to {submission_prefix}')
zhangwenwei's avatar
zhangwenwei committed
424
            for i, anno in enumerate(det_annos):
zhangwenwei's avatar
zhangwenwei committed
425
                sample_idx = self.data_infos[i]['image']['image_idx']
426
                cur_det_file = f'{submission_prefix}/{sample_idx:06d}.txt'
zhangwenwei's avatar
zhangwenwei committed
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
                with open(cur_det_file, 'w') as f:
                    bbox = anno['bbox']
                    loc = anno['location']
                    dims = anno['dimensions'][::-1]  # lhw -> hwl
                    for idx in range(len(bbox)):
                        print(
                            '{} -1 -1 {:4f} {:4f} {:4f} {:4f} {:4f} {:4f} '
                            '{:4f} {:4f} {:4f} {:4f} {:4f} {:4f} {:4f}'.format(
                                anno['name'][idx],
                                anno['alpha'][idx],
                                *bbox[idx],  # 4 float
                                *dims[idx],  # 3 float
                                *loc[idx],  # 3 float
                                anno['rotation_y'][idx],
                                anno['score'][idx]),
                            file=f,
                        )
444
            print('Result is saved to {}'.format(submission_prefix))
zhangwenwei's avatar
zhangwenwei committed
445
446
447
448
449

        return det_annos

    def convert_valid_bboxes(self, box_dict, info):
        # TODO: refactor this function
450
451
452
        box_preds = box_dict['boxes_3d']
        scores = box_dict['scores_3d']
        labels = box_dict['labels_3d']
zhangwenwei's avatar
zhangwenwei committed
453
        sample_idx = info['image']['image_idx']
454
455
456
        # TODO: remove the hack of yaw
        box_preds.tensor[:, -1] = box_preds.tensor[:, -1] - np.pi
        box_preds.limit_yaw(offset=0.5, period=np.pi * 2)
zhangwenwei's avatar
zhangwenwei committed
457

458
        if len(box_preds) == 0:
zhangwenwei's avatar
zhangwenwei committed
459
            return dict(
460
461
462
463
464
465
                bbox=np.zeros([0, 4]),
                box3d_camera=np.zeros([0, 7]),
                box3d_lidar=np.zeros([0, 7]),
                scores=np.zeros([0]),
                label_preds=np.zeros([0, 4]),
                sample_idx=sample_idx)
zhangwenwei's avatar
zhangwenwei committed
466
467
468
469
470

        rect = info['calib']['R0_rect'].astype(np.float32)
        Trv2c = info['calib']['Tr_velo_to_cam'].astype(np.float32)
        P2 = info['calib']['P2'].astype(np.float32)
        img_shape = info['image']['image_shape']
471
472
473
474
475
        P2 = box_preds.tensor.new_tensor(P2)

        box_preds_camera = box_preds.convert_to(Box3DMode.CAM, rect @ Trv2c)

        box_corners = box_preds_camera.corners
zhangwenwei's avatar
zhangwenwei committed
476
        box_corners_in_image = points_cam2img(box_corners, P2)
zhangwenwei's avatar
zhangwenwei committed
477
478
479
480
481
        # box_corners_in_image: [N, 8, 2]
        minxy = torch.min(box_corners_in_image, dim=1)[0]
        maxxy = torch.max(box_corners_in_image, dim=1)[0]
        box_2d_preds = torch.cat([minxy, maxxy], dim=1)
        # Post-processing
482
483
484
485
486
487
488
489
490
491
        # check box_preds_camera
        image_shape = box_preds.tensor.new_tensor(img_shape)
        valid_cam_inds = ((box_preds_camera.tensor[:, 0] < image_shape[1]) &
                          (box_preds_camera.tensor[:, 1] < image_shape[0]) &
                          (box_preds_camera.tensor[:, 2] > 0) &
                          (box_preds_camera.tensor[:, 3] > 0))
        # check box_preds
        limit_range = box_preds.tensor.new_tensor(self.pcd_limit_range)
        valid_pcd_inds = ((box_preds.center > limit_range[:3]) &
                          (box_preds.center < limit_range[3:]))
zhangwenwei's avatar
zhangwenwei committed
492
493
494
495
496
        valid_inds = valid_cam_inds & valid_pcd_inds.all(-1)

        if valid_inds.sum() > 0:
            return dict(
                bbox=box_2d_preds[valid_inds, :].numpy(),
497
498
499
500
                box3d_camera=box_preds_camera[valid_inds].tensor.numpy(),
                box3d_lidar=box_preds[valid_inds].tensor.numpy(),
                scores=scores[valid_inds].numpy(),
                label_preds=labels[valid_inds].numpy(),
zhangwenwei's avatar
zhangwenwei committed
501
502
503
504
                sample_idx=sample_idx,
            )
        else:
            return dict(
505
506
507
508
509
                bbox=np.zeros([0, 4]),
                box3d_camera=np.zeros([0, 7]),
                box3d_lidar=np.zeros([0, 7]),
                scores=np.zeros([0]),
                label_preds=np.zeros([0, 4]),
zhangwenwei's avatar
zhangwenwei committed
510
511
                sample_idx=sample_idx,
            )