-
Notifications
You must be signed in to change notification settings - Fork 2
/
visualization_from_json.py
267 lines (222 loc) · 10.8 KB
/
visualization_from_json.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
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
import argparse
import json
import os
import cv2
import matplotlib.pyplot as plt
import numpy as np
import torch
from PIL import Image
from matplotlib.figure import Figure
from tqdm import trange
from e2plabel.e2plabelconvert import generatePerspective, VIEW_NAME, VIEW_ARGS
from postprocess.postprocess2 import _cal_p_pred_emask
from visualization import clearAxesLines
img_hw = (512, 512)
e_img_hw = (512, 1024)
def jsonToCor(filename):
H, W = e_img_hw
with open(filename) as f:
inferenced_result = json.load(f)
cor_id = np.array(inferenced_result['uv'], np.float32)
cor_id[:, 0] *= W
cor_id[:, 1] *= H
return cor_id
def txtToCor(filename):
with open(filename) as f:
cor = np.array([line.strip().split() for line in f if line.strip()], np.float32)
return cor
def resolveImgPath(s: str):
if s.find("pano") == 0 or s.find("camera") == 0:
return os.path.join("data/layoutnet_dataset/test/img", s)
else:
return os.path.join("data/matterport3d_layout/test/img", s)
def resolveGtCorPath(s: str):
if s.find("pano") == 0 or s.find("camera") == 0:
return os.path.join("data/layoutnet_dataset/test/label_cor", os.path.splitext(s)[0] + ".txt")
else:
return os.path.join("data/matterport3d_layout/test/label_cor", os.path.splitext(s)[0] + ".txt")
def corTo2DMask(e_img, cor):
pres = generatePerspective(e_img, cor, VIEW_NAME, VIEW_ARGS, img_hw)
lines = []
for d in pres:
lines.append(torch.tensor(d["lines"]))
masks2d = []
for view_idx in range(6):
thickness = int(round(img_hw[0] * 0.01))
mat = np.zeros((1, *img_hw))
for line in lines[view_idx]:
cv2.line(mat[0], torch.round(line[3:5]).to(torch.int64).numpy(),
torch.round(line[5:7]).to(torch.int64).numpy(), 1.0, thickness=thickness)
masks2d.append(torch.tensor(mat))
masks2d = torch.stack(masks2d)
maskEq = _cal_p_pred_emask(None, masks2d, img_hw, e_img_hw)
return maskEq
def wireframeGetMaskImg(e_img, cor, color) -> torch.Tensor:
maskEq = corTo2DMask(e_img, cor).squeeze()
mask_img = torch.cat([torch.tensor(color).repeat(*maskEq.shape[0:2], 1), maskEq.unsqueeze(-1)], 2)
mask_img = torch.round(mask_img * 255).to(torch.uint8)
return mask_img
def drawWireframeOnEImg(e_img, cor, color):
plt.imshow(wireframeGetMaskImg(e_img, cor, color).cpu().numpy())
fig: Figure = None
def show(output_path, name):
if output_path:
plt.savefig(os.path.join(output_path, imgPath + "." + name + ".png"))
else:
plt.show()
plt.close(fig)
def initFig():
global fig
fig = plt.figure(figsize=(10.24, 5.12))
plt.gcf().subplots_adjust(top=1, bottom=0, left=0, right=1, hspace=0, wspace=0)
ax = plt.gca()
clearAxesLines(ax)
CLASS_A = [
"7y3sRwLe3Va_9b72664399a34e4f9dbe470571c73187.png",
"B6ByNegPMKs_8b1abc1b47784d758b9ec1e079160475.png",
"camera_1a2b3c7901434d88bba55d6f2b28a6d5_office_30_frame_equirectangular_domain_.png",
"camera_7a42df17b40c4c15bfd6301823b6a476_office_22_frame_equirectangular_domain_.png",
"camera_8cbbb3e42c0e4e54b3b523b1fec6b3bc_office_33_frame_equirectangular_domain_.png",
"camera_412ba0d035b5432abd88ed447716f349_office_30_frame_equirectangular_domain_.png",
"camera_514bd77b98cc47ad904d6c8196f769b1_office_8_frame_equirectangular_domain_.png",
"camera_d162082c8f714aee8984195e0c5a7396_office_11_frame_equirectangular_domain_.png",
"e9zR4mvMWw7_f624a40d100144e696a39abe258ee090.png",
"pano_adxsvoaiehisue.png",
"pano_agpqpoljoyzxds.png",
"pano_ahvuapixtvirde.png",
"pano_aixninerbhvojf.png",
"pano_ankughvvgbhsom.png",
"pano_apozlylyjgtjid.png",
"uNb9QFRL6hY_1434b965c3c147419c4ff40310633b58.png",
"x8F5xyUWy9e_2669f5ba693c4e729d7d2c4f3fa0a077.png",
]
CLASS_B = [
"pano_aghlgnaxvjlzmb.png",
"7y3sRwLe3Va_92fb09a83f8949619b9dc5bda2855456.png",
"7y3sRwLe3Va_fdab6422162e49db822a37178ab70481.png",
"B6ByNegPMKs_53249ef8a94c4c40bd6f09c069e54d16.png",
"B6ByNegPMKs_bb2332e3d7ad40a59ee5ad0eae108dec.png",
"B6ByNegPMKs_ce2f5a74556c4be192df3ca7a178cefb.png",
"camera_32caf5752a4746c8b95f84e9acd9271d_office_29_frame_equirectangular_domain_.png",
"camera_63eb2cd447b84c5abac846f79c51dfcd_office_14_frame_equirectangular_domain_.png",
"camera_90af0a7fe0ed4a7db2c2e05727560231_office_15_frame_equirectangular_domain_.png",
"camera_270448008f5743f48f34539d36e4c4ae_office_14_frame_equirectangular_domain_.png",
"pano_auqcjiehbmenao.png",
"wc2JMjhGNzB_6e491bc8576345bda3cdde9ab216b7be.png",
]
CLASS_C_D = [
"7y3sRwLe3Va_9e4c92fd7eb74504baecf55a3264716e.png",
"7y3sRwLe3Va_6376b741b50a4418b3dc3fde791c3c09.png",
"B6ByNegPMKs_5b3d1c9fefb64512b0c9750a00feece4.png",
"B6ByNegPMKs_e5567bd5fa2d4fde8a6b9f15e3274a7e.png",
"e9zR4mvMWw7_5d711de78dbd400aa4cfd51fc05dfbee.png",
"pano_abbvryjplnajxo.png",
"pano_aqdafdzfhdukpg.png",
"uNb9QFRL6hY_d11f14ddecbe406681d4980365ea5a43.png",
"7y3sRwLe3Va_dd83fb40a2e14ac99de9fe9bcfaf44df.png",
"uNb9QFRL6hY_bcce4f23c12744c782c0b49b24a0331a.png",
"camera_a39f4a868cd84429a765324af21c6e6e_office_8_frame_equirectangular_domain_.png",
]
PANO_ARR = []
STF_ARR = []
MATTER_ARR = []
if __name__ == '__main__':
parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--img', help='指定一张图片。如果不指定，那就会画所有')
parser.add_argument('--output_path', help='如果不指定，就会plt.show')
parser.add_argument('--draw_independent', "-d", action="store_true", help='独立画图还是一张图画好几次？')
parser.add_argument('--draw_both', "-b", action="store_true", help='两种方法都画')
parser.add_argument('--second', "-2", action="store_true", help='定义此项则画HoHoNet和AtlantaNet，否则画HorizonNet和LayoutNet')
args = parser.parse_args()
GT_COLOR = (0.0, 1.0, 0.0) # 绿
OUR_PATH, OUR_COLOR = "result_json", (1.0, 0.0, 0.0) # 红
HORIZONNET_PATH, HORIZONNET_COLOR = "eval_results/horizonnet_json", (0.0, 0.0, 1.0) # 蓝
LAYOUTNET_PATH, LAYOUTNET_COLOR = "eval_results/layoutnet_json", (1.0, 0.0, 1.0) # 粉
ATLANTANET_PATH, ATLANTANET_COLOR = "eval_results/atlantanet_json", (0.0, 0.0, 1.0) # 蓝
HOHONET_PATH, HOHONET_COLOR = "eval_results/hohonet_json", (1.0, 0.0, 1.0) # 粉
if args.img:
img_list = [args.img]
else:
img_list = [s.replace(".json", "") for s in os.listdir(OUR_PATH)]
# # TODO
# img_list = CLASS_A + CLASS_B + CLASS_C_D
# args.output_path = "result_6_pick"
if args.output_path:
os.makedirs(args.output_path, exist_ok=True)
for i in trange(len(img_list)):
imgPath = img_list[i]
isPanoStf = imgPath.find("pano") == 0 or imgPath.find("camera") == 0
if args.second and isPanoStf: continue
e_img = np.array(Image.open(resolveImgPath(imgPath))) / 255.0
gt_cor = txtToCor(resolveGtCorPath(imgPath))
myJsonPath = os.path.join(OUR_PATH, imgPath + ".json")
with open(myJsonPath) as f:
my_result = json.load(f)
iou3d = my_result["3DIoU"]
# # TODO
# if imgPath.find("pano") == 0:
# PANO_ARR.append(iou3d)
# elif imgPath.find("camera") == 0:
# STF_ARR.append(iou3d)
# else:
# MATTER_ARR.append(iou3d)
# continue
# 画的顺序：gt、layout、horizon、ours
if args.draw_both or (not args.draw_independent):
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
if not args.second:
drawWireframeOnEImg(e_img, txtToCor(os.path.join(LAYOUTNET_PATH, os.path.splitext(imgPath)[0] + (
"_aligned_rgb" if isPanoStf else "") + "_cor_id.txt")), LAYOUTNET_COLOR)
drawWireframeOnEImg(e_img, jsonToCor(os.path.join(HORIZONNET_PATH, os.path.splitext(imgPath)[0] + ".json")),
HORIZONNET_COLOR)
else:
drawWireframeOnEImg(e_img, jsonToCor(os.path.join(ATLANTANET_PATH, os.path.splitext(imgPath)[0] + ".json")),
ATLANTANET_COLOR)
drawWireframeOnEImg(e_img, txtToCor(os.path.join(HOHONET_PATH, os.path.splitext(imgPath)[0] + ".layout.txt")),
HOHONET_COLOR)
drawWireframeOnEImg(e_img, jsonToCor(myJsonPath), OUR_COLOR)
show(args.output_path, "all.{:.2f}".format(iou3d))
if args.draw_both or args.draw_independent:
if not args.second:
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
drawWireframeOnEImg(e_img, txtToCor(os.path.join(LAYOUTNET_PATH, os.path.splitext(imgPath)[0] + (
"_aligned_rgb" if isPanoStf else "") + "_cor_id.txt")), LAYOUTNET_COLOR)
show(args.output_path, "lay")
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
drawWireframeOnEImg(e_img, jsonToCor(os.path.join(HORIZONNET_PATH, os.path.splitext(imgPath)[0] + ".json")),
HORIZONNET_COLOR)
show(args.output_path, "hor")
else:
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
drawWireframeOnEImg(e_img, jsonToCor(os.path.join(ATLANTANET_PATH, os.path.splitext(imgPath)[0] + ".json")),
ATLANTANET_COLOR)
show(args.output_path, "atl".format(iou3d))
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
drawWireframeOnEImg(e_img, txtToCor(os.path.join(HOHONET_PATH, os.path.splitext(imgPath)[0] + ".layout.txt")),
HOHONET_COLOR)
show(args.output_path, "hoh".format(iou3d))
initFig()
plt.imshow(e_img)
drawWireframeOnEImg(e_img, gt_cor, GT_COLOR)
drawWireframeOnEImg(e_img, jsonToCor(myJsonPath), OUR_COLOR)
show(args.output_path, "our.{:.2f}".format(iou3d))
a = 1
# TODO
# import torch
# PANO_ARR = torch.tensor(PANO_ARR).sort(descending=True)[0]
# STF_ARR = torch.tensor(STF_ARR).sort(descending=True)[0]
# MATTER_ARR = torch.tensor(MATTER_ARR).sort(descending=True)[0]
# for a in [PANO_ARR,STF_ARR,MATTER_ARR]:
# pt = [round(len(a) / 4 * (i+1)) for i in range(3)]
# pt = [a[v] for v in pt]
# print(pt)