-
Notifications
You must be signed in to change notification settings - Fork 10
/
Copy pathpcdvis.py
154 lines (122 loc) · 5.51 KB
/
pcdvis.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
import open3d
import argparse
import os
import time
import json
import h5py
import datetime
import cv2
import yaml
import colorsys
import numpy as np
import pandas as pd
import torch
from torch.utils.data import DataLoader
import torch.nn.functional as F
from tqdm import tqdm
from matplotlib import pyplot as plt
import my_log as log
from model.pointnet import PointNetSeg, feature_transform_reguliarzer
from model.pointnet2 import PointNet2SemSeg
from model.utils import load_pointnet
from pcdseg import parse_args
from data_utils.SemKITTI_Loader import pcd_normalize
from data_utils.kitti_utils import Semantic_KITTI_Utils
KITTI_ROOT = os.environ['KITTI_ROOT']
class Window_Manager():
def __init__(self):
self.param = open3d.io.read_pinhole_camera_parameters('config/ego_view.json')
self.vis = open3d.visualization.VisualizerWithKeyCallback()
self.vis.create_window(width=800, height=800, left=100)
self.vis.register_key_callback(32, lambda vis: exit())
self.vis.get_render_option().load_from_json('config/render_option.json')
self.pcd = open3d.geometry.PointCloud()
def update(self, pts_3d, colors):
self.pcd.points = open3d.utility.Vector3dVector(pts_3d)
self.pcd.colors = open3d.utility.Vector3dVector(colors/255)
self.vis.remove_geometry(self.pcd)
self.vis.add_geometry(self.pcd)
self.vis.get_view_control().convert_from_pinhole_camera_parameters(self.param)
self.vis.update_geometry(self.pcd)
self.vis.poll_events()
self.vis.update_renderer()
def capture_screen(self,fn):
self.vis.capture_screen_image(fn, False)
def export_video():
fourcc = cv2.VideoWriter_fourcc(*'XVID')
font = cv2.FONT_HERSHEY_SIMPLEX
out = cv2.VideoWriter('experiment/pn_compare.avi',fourcc, 15.0, (int(1600*0.8),int(740*0.8)))
# mkdir('experiment/imgs/%s/'%(args.model_name))
# vis_handle.capture_screen('experiment/imgs/%s/%d_3d.png'%(args.model_name,i))
# cv2.imwrite('experiment/imgs/%s/%d_sem.png'%(args.model_name, i), img_semantic)
for index in range(100, 320):
pn_3d = cv2.imread('experiment/imgs/pointnet/%d_3d.png' % (index))
pn_sem = cv2.imread('experiment/imgs/pointnet/%d_sem.png' % (index))
pn2_3d = cv2.imread('experiment/imgs/pointnet2/%d_3d.png' % (index))
pn2_sem = cv2.imread('experiment/imgs/pointnet2/%d_sem.png' % (index))
pn_3d = pn_3d[160:650]
pn2_3d = pn2_3d[160:650]
pn_sem = cv2.resize(pn_sem, (800, 250))
pn2_sem = cv2.resize(pn2_sem, (800, 250))
pn = np.vstack((pn_3d, pn_sem))
pn2 = np.vstack((pn2_3d, pn2_sem))
cv2.putText(pn, 'PointNet', (20, 100), font,1, (255, 255, 255), 2, cv2.LINE_AA)
cv2.putText(pn, 'PointNet', (20, 520), font,1, (255, 255, 255), 2, cv2.LINE_AA)
cv2.putText(pn2, 'PointNet2', (20, 100), font,1, (255, 255, 255), 2, cv2.LINE_AA)
cv2.putText(pn2, 'PointNet2', (20, 520), font,1, (255, 255, 255), 2, cv2.LINE_AA)
merge = np.hstack((pn, pn2))
class_names = ['unlabelled', 'vehicle', 'human', 'ground', 'structure', 'nature']
colors = [[255, 255, 255],[245, 150, 100],[30, 30, 255],[255, 0, 255],[0, 200, 255],[0, 175, 0]]
for i,(name,c) in enumerate(zip(class_names, colors)):
cv2.putText(merge, name, (200 + i * 200, 50), font,1, [c[2],c[1],c[0]], 2, cv2.LINE_AA)
cv2.line(merge,(0,70),(1600,70),(255,255,255),2)
cv2.line(merge,(800,70),(800,1300),(255,255,255),2)
merge = cv2.resize(merge,(0,0),fx=0.8,fy=0.8)
# cv2.imshow('merge', merge)
# if 32 == waitKey(1):
# break
out.write(merge)
print(index)
out.release()
def vis(args):
part = '01'
args.subset ='inview'
args.model_name = 'pointnet'
kitti_utils = Semantic_KITTI_Utils(KITTI_ROOT, subset=args.subset)
vis_handle = Window_Manager()
if args.model_name == 'pointnet':
args.pretrain = 'checkpoints/pointnet-inview-0.52077-0018.pth'
else:
args.pretrain = 'checkpoints/pointnet2-inview-0.55884-0001.pth'
model = load_pointnet(args.model_name, kitti_utils.num_classes, args.pretrain)
for index in range(0, kitti_utils.get_max_index(part)):
point_cloud, label = kitti_utils.get(part, index, load_image=True)
# resample point cloud
length = point_cloud.shape[0]
npoints = 25000
choice = np.random.choice(length, npoints, replace=True)
point_cloud = point_cloud[choice]
label = label[choice]
pts_3d = point_cloud[:,:3]
pcd = pcd_normalize(point_cloud)
with log.Tick():
points = torch.from_numpy(pcd).unsqueeze(0).transpose(2, 1).cuda()
with torch.no_grad():
if args.model_name == 'pointnet':
logits, _ = model(points)
else:
logits = model(points)
pred = logits[0].argmax(-1).cpu().numpy()
print(index, pred.shape, end='')
# pts_2d = kitti_utils.project_3d_to_2d(pts_3d)
pts_2d = kitti_utils.torch_project_3d_to_2d(pts_3d)
vis_handle.update(pts_3d, kitti_utils.colors[pred])
sem_img = kitti_utils.draw_2d_points(pts_2d, kitti_utils.colors_bgr[pred])
cv2.imshow('semantic', sem_img)
cv2.imshow('frame', cv2.cvtColor(kitti_utils.frame,cv2.COLOR_BGR2RGB))
if 32 == cv2.waitKey(1):
break
if __name__ == '__main__':
args = parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
vis(args)