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
|
from torch.utils import data
from typing import Tuple, Union, List
import numpy as np
import json
import math
import cv2
import h5py
import os
import pickle
class RamImage():
def __init__(self, path):
fd = open(path, 'rb')
img_str = fd.read()
fd.close()
self.img_raw = np.frombuffer(img_str, np.uint8)
def to_numpy(self):
return cv2.imdecode(self.img_raw, cv2.IMREAD_COLOR)
class ClevrerSample(data.Dataset):
def __init__(self, root_path: str, data_path: str, size: Tuple[int, int]):
self.size = size
self.data_path = root_path
self.video_id = int(data_path.split('_')[1])
frames = []
for frame in os.listdir(os.path.join(root_path, data_path)):
if os.path.isfile(os.path.join(root_path, data_path, frame)) and frame.endswith('.jpg'):
frames.append(os.path.join(root_path, data_path, frame))
frames.sort()
self.imgs = []
for path in frames:
self.imgs.append(RamImage(path))
def get_data(self):
frames = np.zeros((128,3,self.size[1], self.size[0]),dtype=np.float32)
for i in range(len(self.imgs)):
img = self.imgs[i].to_numpy()
frames[i] = img.transpose(2, 0, 1).astype(np.float32) / 255.0
return frames
def downsample(self, size):
self.size = size
imgs = []
path = os.path.join(self.data_path, 'tmp.jpg')
for image_large in self.imgs:
img_small = cv2.resize(image_large.to_numpy(), dsize=(self.size[0], self.size[1]), interpolation=cv2.INTER_CUBIC)
cv2.imwrite(path, img_small)
imgs.append(RamImage(path))
self.imgs = imgs
# remove tmp.jpg
os.remove(path)
return self
class ClevrerDataset(data.Dataset):
def save(self):
with open(self.file, "wb") as outfile:
pickle.dump(self.samples, outfile)
def load(self):
with open(self.file, "rb") as infile:
self.samples = pickle.load(infile)
def __init__(self, root_path: str, dataset_name: str, type: str, size: Tuple[int, int], full_size: Tuple[int, int] = None, use_slotformer: bool = False, evaluation: bool = False):
data_path = f'data/data/video/{dataset_name}'
data_path = os.path.join(root_path, data_path)
self.file = os.path.join(data_path, f'dataset-{size[0]}x{size[1]}-{type}.pickle')
self.samples = []
if os.path.exists(self.file):
self.load()
else:
if (full_size is None) or (size == full_size):
sample_path = os.path.join(data_path, type, 'images')
samples = []
print(f'sample_path:', sample_path)
for directory in os.listdir(sample_path):
path = os.path.join(sample_path, directory)
if os.path.isdir(path):
samples.append(directory)
samples.sort()
num_samples = len(samples)
print(f'num_samples:', num_samples)
for i, dir in enumerate(samples):
self.samples.append(ClevrerSample(sample_path, dir, size))
print(f"Loading CLEVRER [{i * 100 / num_samples:.2f}]", flush=True)
else:
# load full size dataset
full_dataset = ClevrerDataset(root_path, dataset_name, type, full_size, full_size)
# downsample
for i, sample in enumerate(full_dataset.samples):
self.samples.append(sample.downsample(size))
print(f"Loading CLEVRER {type} [{i * 100 / len(full_dataset.samples):.2f}]", flush=True)
self.save()
self.length = len(self.samples)
self.background = None
if "background.jpg" in os.listdir(data_path):
self.background = cv2.imread(os.path.join(data_path, "background.jpg"))
self.background = cv2.resize(self.background, dsize=size, interpolation=cv2.INTER_CUBIC)
self.background = self.background.transpose(2, 0, 1).astype(np.float32) / 255.0
self.background = self.background.reshape(self.background.shape[0], self.background.shape[1], self.background.shape[2])
self.use_slotformer = use_slotformer
self.eval = evaluation
if self.use_slotformer:
with open(f'{data_path}/slotformer/valid_idx_{type}.pt', 'rb') as f:
self.slotformer_idx = pickle.load(f)
self.length = len(self.slotformer_idx)
self.video_ids = [sample.video_id for sample in self.samples]
self.burn_in_length = 6
self.rollout_length = 10
self.skip_length = 2
if self.eval:
self.gt_mask = np.load(f'{data_path}/slotformer/gt_mask.npy').astype(np.int8)
self.gt_bbox = np.load(f'{data_path}/slotformer/gt_bbox.npy').astype(np.int16)
self.gt_pres_mask = np.load(f'{data_path}/slotformer/gt_pres_mask.npy').astype(bool)
#self.gt = np.load(f'{data_path}/slotformer/gt.npy').astype(np.float32)
print(f"ClevrerDataset: {self.length}")
if len(self) == 0:
raise FileNotFoundError(f'Found no dataset at {self.data_path}')
if False:
for sample in self.samples:
frames = sample.get_data()
i = 0
for frame in frames:
i += 1
if i == 29:
print('test')
frame = np.einsum('chw->hwc', frame)
cv2.imshow('frame', frame)
cv2.waitKey(0)
def __len__(self):
return self.length
def __getitem__(self, index: int):
if self.use_slotformer:
index_org = index
_, start_idx, video_id = self.slotformer_idx[index]
index = self.video_ids.index(video_id)
#print(f"Loading CLEVRER {index} {start_idx} {video_id}")
if video_id != self.samples[index].video_id:
raise ValueError(f"Video ID mismatch {video_id} != {self.samples[index].video_id}")
selec = range(start_idx, start_idx+(self.burn_in_length+self.rollout_length)*self.skip_length, self.skip_length)
frames = self.samples[index].get_data()[selec]
if self.eval:
gt_mask = self.gt_mask[index_org]
gt_bbox = self.gt_bbox[index_org]
gt_pres_mask = self.gt_pres_mask[index_org]
#gt = self.gt[index_org]
return (
frames,
self.background,
gt_mask,
gt_bbox,
gt_pres_mask,
#gt
)
else:
frames = self.samples[index].get_data()
return (
frames,
self.background
)
#if __name__ == "__main__":
#dataset = ClevrerDataset('./', "CLEVRER", "train", [320, 240])
#dataset = ClevrerDataset('./', "CLEVRER", "train", [480, 320])
#dataset = ClevrerDataset('./', "CLEVRER", "train", [120, 80], [480, 320])
|
