-
Notifications
You must be signed in to change notification settings - Fork 18
/
visualise.py
153 lines (127 loc) · 5.85 KB
/
visualise.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
# ------------------------------------------------------------------------
# PowerBEV
# Copyright (c) 2023 Peizheng Li. All Rights Reserved.
# ------------------------------------------------------------------------
# Modified from FIERY (https://github.com/wayveai/fiery)
# Copyright (c) 2021 Wayve Technologies Limited. All Rights Reserved.
# ------------------------------------------------------------------------
import os
from argparse import ArgumentParser
from glob import glob
import cv2
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import torch
import torchvision
from PIL import Image
from powerbev.config import get_cfg, get_parser
from powerbev.data import prepare_powerbev_dataloaders
from powerbev.trainer import TrainingModule
from powerbev.utils.instance import predict_instance_segmentation, generate_gt_instance_segmentation
from powerbev.utils.network import NormalizeInverse
from powerbev.utils.visualisation import (convert_figure_numpy,
generate_instance_colours,
make_contour, plot_instance_map)
def plot_prediction(image, output, cfg):
if cfg.VISUALIZATION.VIS_GT:
# Process ground truth
consistent_instance_seg, matched_centers = generate_gt_instance_segmentation(
output, compute_matched_centers=True, spatial_extent=(cfg.LIFT.X_BOUND[1], cfg.LIFT.Y_BOUND[1])
)
else:
# Process predictions
consistent_instance_seg, matched_centers = predict_instance_segmentation(
output, compute_matched_centers=True, spatial_extent=(cfg.LIFT.X_BOUND[1], cfg.LIFT.Y_BOUND[1])
)
first_instance_seg = consistent_instance_seg[0, 1]
# Plot future trajectories
unique_ids = torch.unique(first_instance_seg).cpu().long().numpy()[1:]
instance_map = dict(zip(unique_ids, unique_ids))
instance_colours = generate_instance_colours(instance_map)
vis_image = plot_instance_map(first_instance_seg.cpu().numpy(), instance_map)
trajectory_img = np.zeros(vis_image.shape, dtype=np.uint8)
for instance_id in unique_ids:
path = matched_centers[instance_id]
for t in range(len(path) - 1):
color = instance_colours[instance_id].tolist()
cv2.line(trajectory_img, tuple(path[t]), tuple(path[t + 1]),
color, 4)
# Overlay arrows
temp_img = cv2.addWeighted(vis_image, 0.7, trajectory_img, 0.3, 1.0)
mask = ~ np.all(trajectory_img == 0, axis=2)
vis_image[mask] = temp_img[mask]
# Plot present RGB frames and predictions
val_w = 2.99
cameras = cfg.IMAGE.NAMES
image_ratio = cfg.IMAGE.FINAL_DIM[0] / cfg.IMAGE.FINAL_DIM[1]
val_h = val_w * image_ratio
fig = plt.figure(figsize=(4 * val_w, 2 * val_h))
width_ratios = (val_w, val_w, val_w, val_w)
gs = mpl.gridspec.GridSpec(2, 4, width_ratios=width_ratios)
gs.update(wspace=0.0, hspace=0.0, left=0.0, right=1.0, top=1.0, bottom=0.0)
denormalise_img = torchvision.transforms.Compose(
(NormalizeInverse(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
torchvision.transforms.ToPILImage(),)
)
for imgi, img in enumerate(image[0, -1]):
ax = plt.subplot(gs[imgi // 3, imgi % 3])
showimg = denormalise_img(img.cpu())
if imgi > 2:
showimg = showimg.transpose(Image.FLIP_LEFT_RIGHT)
plt.annotate(cameras[imgi].replace('_', ' ').replace('CAM ', ''), (0.01, 0.87), c='white',
xycoords='axes fraction', fontsize=14)
plt.imshow(showimg)
plt.axis('off')
ax = plt.subplot(gs[:, 3])
plt.imshow(make_contour(vis_image[::-1, ::-1]))
plt.axis('off')
plt.draw()
figure_numpy = convert_figure_numpy(fig)
plt.close()
return figure_numpy
def visualise():
args = get_parser().parse_args()
cfg = get_cfg(args)
_, valloader = prepare_powerbev_dataloaders(cfg)
trainer = TrainingModule(cfg.convert_to_dict())
if cfg.PRETRAINED.LOAD_WEIGHTS:
# Load single-image instance segmentation model.
weights_path = cfg.PRETRAINED.PATH
pretrained_model_weights = torch.load(
weights_path , map_location='cpu'
)['state_dict']
trainer.load_state_dict(pretrained_model_weights, strict=False)
print(f'Loaded single-image model weights from {weights_path}')
device = torch.device('cuda:0')
trainer = trainer.to(device)
trainer.eval()
for i, batch in enumerate(valloader):
if cfg.VISUALIZATION.VIS_GT:
# Visualize ground truth
image = batch['image'].to(device)
time_range = cfg.TIME_RECEPTIVE_FIELD - 2
output = {
'segmentation': batch['segmentation'][:, time_range:].to(device),
'instance_flow': batch['flow'][:, time_range:].to(device),
'centerness': batch['centerness'][:, time_range:].to(device),
}
figure_numpy = plot_prediction(image, output, cfg)
else:
# Visualize predictions
image = batch['image'].to(device)
intrinsics = batch['intrinsics'].to(device)
extrinsics = batch['extrinsics'].to(device)
future_egomotions = batch['future_egomotion'].to(device)
# Forward pass
with torch.no_grad():
output = trainer.model(image, intrinsics, extrinsics, future_egomotions)
figure_numpy = plot_prediction(image, output, cfg)
os.makedirs(os.path.join(cfg.VISUALIZATION.OUTPUT_PATH), exist_ok=True)
output_filename = os.path.join(cfg.VISUALIZATION.OUTPUT_PATH, 'sample_'+str(i)) + '.png'
Image.fromarray(figure_numpy).save(output_filename)
print(f'Saved output in {output_filename}')
if i >= cfg.VISUALIZATION.SAMPLE_NUMBER-1:
return
if __name__ == '__main__':
visualise()