-
Notifications
You must be signed in to change notification settings - Fork 0
/
pc_fusion.py
190 lines (153 loc) · 7.54 KB
/
pc_fusion.py
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
"""
Fuses depth maps into point clouds using the PC fuser from
https://github.com/alexrich021/3dvnet/blob/main/mv3d/eval/pointcloudfusion_custom.py
This script follows the format in test.py. It expects a model to use for
depth prediction.
Example command:
python pc_fusion.py --name HERO_MODEL \
--output_base_path OUTPUT_PATH \
--config_file models/hero_model.yaml \
--load_weights_from_checkpoint models/hero_model.ckpt \
--data_config configs/data/scannet_default_test.yaml \
--num_workers 8 \
--batch_size 8;
"""
import os
from pathlib import Path
import open3d as o3d
import torch
import torch.nn.functional as F
from tqdm import tqdm
from experiment_modules.depth_model import DepthModel
import options
import tools.torch_point_cloud_fusion as torch_point_cloud_fusion
from utils.dataset_utils import get_dataset
from utils.generic_utils import (to_gpu, reverse_imagenet_normalize)
import modules.cost_volume as cost_volume
def main(opts):
# get dataset
dataset_class, scans = get_dataset(opts.dataset,
opts.dataset_scan_split_file, opts.single_debug_scan_id)
# fusion params
N_CONSISTENT_THRESH = opts.n_consistent_thresh
Z_THRESH = opts.pc_fusion_z_thresh
VOXEL_DOWNSAMPLE = opts.voxel_downsample
# output location
pc_output_folder_name = f"{N_CONSISTENT_THRESH}_{Z_THRESH}_{VOXEL_DOWNSAMPLE}_{opts.fusion_max_depth}"
# path where results for this model, dataset, and tuple type are.
results_path = os.path.join(opts.output_base_path, opts.name,
opts.dataset, opts.frame_tuple_type)
# ouput path
pcs_output_dir = os.path.join(results_path, "pcs", pc_output_folder_name)
Path(os.path.join(pcs_output_dir)).mkdir(parents=True, exist_ok=True)
print(f"".center(80, "#"))
print(f" Running PC Fusion!".center(80, "#"))
print(f"Output directory:\n{pcs_output_dir} ".center(80, "#"))
print(f"".center(80, "#"))
print("")
# load model
model = DepthModel.load_from_checkpoint(
opts.load_weights_from_checkpoint,
args=None)
if (opts.fast_cost_volume and
isinstance(model.cost_volume, cost_volume.FeatureVolumeManager)):
model.cost_volume = model.cost_volume.to_fast()
model = model.cuda().eval()
with torch.inference_mode():
for scan in tqdm(scans):
# set up dataset with current scan
dataset = dataset_class(
opts.dataset_path,
split=opts.split,
mv_tuple_file_suffix=opts.mv_tuple_file_suffix,
limit_to_scan_id=scan,
include_full_res_depth=True,
tuple_info_file_location=opts.tuple_info_file_location,
num_images_in_tuple=None,
shuffle_tuple=opts.shuffle_tuple,
include_high_res_color=True,
include_full_depth_K=True,
skip_frames=opts.skip_frames,
skip_to_frame=opts.skip_to_frame,
image_width=opts.image_width,
image_height=opts.image_height,
pass_frame_id=True,
)
dataloader = torch.utils.data.DataLoader(
dataset,
batch_size=opts.batch_size,
shuffle=False,
num_workers=opts.num_workers,
drop_last=False,
)
# loop and collate data
images_list = []
depths_list = []
poses_list = []
K_list = []
for _, batch in enumerate(tqdm(dataloader)):
# get data, move to GPU
cur_data, src_data = batch
cur_data = to_gpu(cur_data, key_ignores=["frame_id_string"])
src_data = to_gpu(src_data, key_ignores=["frame_id_string"])
outputs = model(
"test",
cur_data, src_data,
unbatched_matching_encoder_forward=True,
return_mask=True,
)
depth_pred_s0_b1hw = outputs["depth_pred_s0_b1hw"].cuda()
depth_pred_s0_b1hw[depth_pred_s0_b1hw >
opts.fusion_max_depth] = 0
upsampled_depth_pred = F.interpolate(
depth_pred_s0_b1hw,
size=(480, 640),
mode="nearest",
)
depths_list.append(upsampled_depth_pred)
poses_list.append(cur_data["cam_T_world_b44"].clone())
K_33 = cur_data["K_s0_b44"].clone()
K_33[:,0] *= (640/depth_pred_s0_b1hw.shape[-1])
K_33[:,1] *= (480/depth_pred_s0_b1hw.shape[-2])
K_list.append(K_33.clone())
cur_data["high_res_color_b3hw"] = F.interpolate(
cur_data["high_res_color_b3hw"],
size=(480, 640),
mode="bilinear",
)
image = cur_data["high_res_color_b3hw"].cuda()
image = reverse_imagenet_normalize(image)
images_list.append(image)
# pass data to pc fuser
depths_preds_bhw = torch.cat(depths_list, dim=0).squeeze(1)
poses_b44 = torch.cat(poses_list, dim=0)
image_bhw3 = torch.cat(images_list, dim=0).permute(0,2,3,1)*255
K_b33 = torch.cat(K_list, dim=0)[:,:3,:3]
fused_pts, fused_rgb, _ = torch_point_cloud_fusion.process_scene(
depths_preds_bhw,
image_bhw3.to(torch.uint8),
poses_b44,
K_b33,
Z_THRESH,
N_CONSISTENT_THRESH,
)
pcd_pred = o3d.geometry.PointCloud()
pcd_pred.points = o3d.utility.Vector3dVector(fused_pts)
pcd_pred.colors = o3d.utility.Vector3dVector(fused_rgb / 255.)
pcd_pred = pcd_pred.voxel_down_sample(VOXEL_DOWNSAMPLE)
pcd_filepath = os.path.join(pcs_output_dir, f"{scan}.ply")
o3d.io.write_point_cloud(pcd_filepath, pcd_pred)
if __name__ == '__main__':
# don't need grad for test.
torch.set_grad_enabled(False)
# get an instance of options and load it with config file(s) and cli args.
option_handler = options.OptionsHandler()
option_handler.parse_and_merge_options()
option_handler.pretty_print_options()
print("\n")
opts = option_handler.options
# if no GPUs are available for us then, use the 32 bit on CPU
if opts.gpus == 0:
print("Setting precision to 32 bits since --gpus is set to 0.")
opts.precision = 32
main(opts)