Skip to content

Commit 3f082fe

Browse files
ssheoreyssheorey
committed
Added Pandaset autonomous driving dataset class and configuration file for RandLANet. (#611)
Fixup to earlier merge commit. Add dataset info (license, citation, website, download, etc.) Style fix
1 parent 16f89cc commit 3f082fe

File tree

1 file changed

+106
-82
lines changed

1 file changed

+106
-82
lines changed

‎ml3d/datasets/pandaset.py

Lines changed: 106 additions & 82 deletions
Original file line numberDiff line numberDiff line change
@@ -1,56 +1,81 @@
11
import os
22
from os.path import join
3-
import numpy as np
4-
import pandas as pd
53
from pathlib import Path
64
import logging
5+
import numpy as np
6+
import pandas as pd
77

88
from .base_dataset import BaseDataset, BaseDatasetSplit
99
from ..utils import make_dir, DATASET
1010

1111
log = logging.getLogger(__name__)
1212

13+
1314
class Pandaset(BaseDataset):
14-
""" This class is used to create a dataset based on the Pandaset autonomous
15+
"""This class is used to create a dataset based on the Pandaset autonomous
1516
driving dataset.
1617
17-
https://pandaset.org/
18-
19-
The dataset includes 42 semantic classes and covers more than 100 scenes,
20-
each of which is 8 seconds long.
21-
18+
PandaSet aims to promote and advance research and development in autonomous
19+
driving and machine learning. The first open-source AV dataset available for
20+
both academic and commercial use, PandaSet combines Hesai’s best-in-class
21+
LiDAR sensors with Scale AI’s high-quality data annotation.
22+
23+
PandaSet features data collected using a forward-facing LiDAR with
24+
image-like resolution (PandarGT) as well as a mechanical spinning LiDAR
25+
(Pandar64). The collected data was annotated with a combination of cuboid
26+
and segmentation annotation (Scale 3D Sensor Fusion Segmentation).
27+
28+
It features::
29+
30+
- 48,000+ camera images
31+
- 16,000+ LiDAR sweeps
32+
- 100+ scenes of 8s each
33+
- 28 annotation classes
34+
- 37 semantic segmentation labels
35+
- Full sensor suite: 1x mechanical spinning LiDAR, 1x forward-facing LiDAR, 6x cameras, On-board GPS/IMU
36+
37+
Website: https://pandaset.org/
38+
Code: https://github.com/scaleapi/pandaset-devkit
39+
Download: https://www.kaggle.com/datasets/usharengaraju/pandaset-dataset/data
40+
Data License: CC0: Public Domain (https://scale.com/legal/pandaset-terms-of-use)
41+
Citation: https://arxiv.org/abs/2112.12610
2242
"""
43+
2344
def __init__(self,
2445
dataset_path,
2546
name="Pandaset",
2647
cache_dir="./logs/cache",
2748
use_cache=False,
2849
ignored_label_inds=[],
2950
test_result_folder='./logs/test_log',
30-
test_split=['115', '116', '117', '119', '120', '124', '139', '149', '158'],
51+
test_split=[
52+
'115', '116', '117', '119', '120', '124', '139', '149',
53+
'158'
54+
],
3155
training_split=[
32-
'001', '002', '003', '005', '011', '013', '015', '016',
33-
'017', '019', '021', '023', '024', '027', '028', '029',
34-
'030', '032', '033', '034', '035', '037', '038', '039',
35-
'040', '041', '042', '043', '044', '046', '052', '053',
36-
'054', '056', '057', '058', '064', '065', '066', '067',
37-
'070', '071', '072', '073', '077', '078', '080', '084',
38-
'088', '089', '090', '094', '095', '097', '098', '101',
39-
'102', '103', '105', '106', '109', '110', '112', '113'
56+
'001', '002', '003', '005', '011', '013', '015', '016',
57+
'017', '019', '021', '023', '024', '027', '028', '029',
58+
'030', '032', '033', '034', '035', '037', '038', '039',
59+
'040', '041', '042', '043', '044', '046', '052', '053',
60+
'054', '056', '057', '058', '064', '065', '066', '067',
61+
'070', '071', '072', '073', '077', '078', '080', '084',
62+
'088', '089', '090', '094', '095', '097', '098', '101',
63+
'102', '103', '105', '106', '109', '110', '112', '113'
64+
],
65+
validation_split=['122', '123'],
66+
all_split=[
67+
'001', '002', '003', '005', '011', '013', '015', '016',
68+
'017', '019', '021', '023', '024', '027', '028', '029',
69+
'030', '032', '033', '034', '035', '037', '038', '039',
70+
'040', '041', '042', '043', '044', '046', '052', '053',
71+
'054', '056', '057', '058', '064', '065', '066', '067',
72+
'069', '070', '071', '072', '073', '077', '078', '080',
73+
'084', '088', '089', '090', '094', '095', '097', '098',
74+
'101', '102', '103', '105', '106', '109', '110', '112',
75+
'113', '115', '116', '117', '119', '120', '122', '123',
76+
'124', '139', '149', '158'
4077
],
41-
validation_split=['122', '123'],
42-
all_split=['001', '002', '003', '005', '011', '013', '015', '016',
43-
'017', '019', '021', '023', '024', '027', '028', '029',
44-
'030', '032', '033', '034', '035', '037', '038', '039',
45-
'040', '041', '042', '043', '044', '046', '052', '053',
46-
'054', '056', '057', '058', '064', '065', '066', '067',
47-
'069', '070', '071', '072', '073', '077', '078', '080',
48-
'084', '088', '089', '090', '094', '095', '097', '098',
49-
'101', '102', '103', '105', '106', '109', '110', '112',
50-
'113', '115', '116', '117', '119', '120', '122', '123',
51-
'124', '139', '149', '158'],
5278
**kwargs):
53-
5479
"""Initialize the function by passing the dataset and other details.
5580
5681
Args:
@@ -79,7 +104,7 @@ def __init__(self,
79104
self.label_to_names = self.get_label_to_names()
80105
self.num_classes = len(self.label_to_names)
81106
self.label_values = np.sort([k for k, v in self.label_to_names.items()])
82-
107+
83108
@staticmethod
84109
def get_label_to_names():
85110
"""Returns a label to names dictionary object.
@@ -89,48 +114,48 @@ def get_label_to_names():
89114
values are the corresponding names.
90115
"""
91116
label_to_names = {
92-
1: "Reflection",
93-
2: "Vegetation",
94-
3: "Ground",
95-
4: "Road",
96-
5: "Lane Line Marking",
97-
6: "Stop Line Marking",
98-
7: "Other Road Marking",
99-
8: "Sidewalk",
100-
9: "Driveway",
101-
10: "Car",
102-
11: "Pickup Truck",
103-
12: "Medium-sized Truck",
104-
13: "Semi-truck",
105-
14: "Towed Object",
106-
15: "Motorcycle",
107-
16: "Other Vehicle - Construction Vehicle",
108-
17: "Other Vehicle - Uncommon",
109-
18: "Other Vehicle - Pedicab",
110-
19: "Emergency Vehicle",
111-
20: "Bus",
112-
21: "Personal Mobility Device",
113-
22: "Motorized Scooter",
114-
23: "Bicycle",
115-
24: "Train",
116-
25: "Trolley",
117-
26: "Tram / Subway",
118-
27: "Pedestrian",
119-
28: "Pedestrian with Object",
120-
29: "Animals - Bird",
121-
30: "Animals - Other",
122-
31: "Pylons",
123-
32: "Road Barriers",
124-
33: "Signs",
125-
34: "Cones",
126-
35: "Construction Signs",
127-
36: "Temporary Construction Barriers",
128-
37: "Rolling Containers",
129-
38: "Building",
117+
1: "Reflection",
118+
2: "Vegetation",
119+
3: "Ground",
120+
4: "Road",
121+
5: "Lane Line Marking",
122+
6: "Stop Line Marking",
123+
7: "Other Road Marking",
124+
8: "Sidewalk",
125+
9: "Driveway",
126+
10: "Car",
127+
11: "Pickup Truck",
128+
12: "Medium-sized Truck",
129+
13: "Semi-truck",
130+
14: "Towed Object",
131+
15: "Motorcycle",
132+
16: "Other Vehicle - Construction Vehicle",
133+
17: "Other Vehicle - Uncommon",
134+
18: "Other Vehicle - Pedicab",
135+
19: "Emergency Vehicle",
136+
20: "Bus",
137+
21: "Personal Mobility Device",
138+
22: "Motorized Scooter",
139+
23: "Bicycle",
140+
24: "Train",
141+
25: "Trolley",
142+
26: "Tram / Subway",
143+
27: "Pedestrian",
144+
28: "Pedestrian with Object",
145+
29: "Animals - Bird",
146+
30: "Animals - Other",
147+
31: "Pylons",
148+
32: "Road Barriers",
149+
33: "Signs",
150+
34: "Cones",
151+
35: "Construction Signs",
152+
36: "Temporary Construction Barriers",
153+
37: "Rolling Containers",
154+
38: "Building",
130155
39: "Other Static Object"
131156
}
132157
return label_to_names
133-
158+
134159
def get_split(self, split):
135160
"""Returns a dataset split.
136161
@@ -142,7 +167,7 @@ def get_split(self, split):
142167
A dataset split object providing the requested subset of the data.
143168
"""
144169
return PandasetSplit(self, split=split)
145-
170+
146171
def get_split_list(self, split):
147172
"""Returns the list of data splits available.
148173
@@ -154,8 +179,8 @@ def get_split_list(self, split):
154179
A dataset split object providing the requested subset of the data.
155180
156181
Raises:
157-
ValueError: Indicates that the split name passed is incorrect. The split name should be one of
158-
'training', 'test', 'validation', or 'all'.
182+
ValueError: Indicates that the split name passed is incorrect. The
183+
split name should be one of 'training', 'test', 'validation', or 'all'.
159184
"""
160185
cfg = self.cfg
161186
dataset_path = cfg.dataset_path
@@ -179,7 +204,7 @@ def get_split_list(self, split):
179204
file_list.append(join(pc_path, f))
180205

181206
return file_list
182-
207+
183208
def is_tested(self, attr):
184209
"""Checks if a datum in the dataset has been tested.
185210
@@ -224,7 +249,7 @@ def save_test_result(self, results, attr):
224249
class PandasetSplit(BaseDatasetSplit):
225250
"""This class is used to create a split for Pandaset dataset.
226251
227-
Args:
252+
Args:
228253
dataset: The dataset to split.
229254
split: A string identifying the dataset split that is usually one of
230255
'training', 'test', 'validation', or 'all'.
@@ -233,6 +258,7 @@ class PandasetSplit(BaseDatasetSplit):
233258
Returns:
234259
A dataset split object providing the requested subset of the data.
235260
"""
261+
236262
def __init__(self, dataset, split='train'):
237263
super().__init__(dataset, split=split)
238264
log.info("Found {} pointclouds for {}".format(len(self.path_list),
@@ -244,19 +270,16 @@ def __len__(self):
244270
def get_data(self, idx):
245271
pc_path = self.path_list[idx]
246272
label_path = pc_path.replace('lidar', 'annotations/semseg')
247-
273+
248274
points = pd.read_pickle(pc_path)
249275
labels = pd.read_pickle(label_path)
250-
276+
251277
intensity = points['i'].to_numpy().astype(np.float32)
252-
points = points.drop(columns=['i', 't', 'd']).to_numpy().astype(np.float32)
278+
points = points.drop(columns=['i', 't', 'd']).to_numpy().astype(
279+
np.float32)
253280
labels = labels.to_numpy().astype(np.int32)
254281

255-
data = {
256-
'point': points,
257-
'intensity': intensity,
258-
'label': labels
259-
}
282+
data = {'point': points, 'intensity': intensity, 'label': labels}
260283

261284
return data
262285

@@ -269,4 +292,5 @@ def get_attr(self, idx):
269292
attr = {'name': name, 'path': pc_path, 'split': self.split}
270293
return attr
271294

295+
272296
DATASET._register_module(Pandaset)

0 commit comments

Comments
 (0)