-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathfinal_results.py
323 lines (254 loc) · 20.9 KB
/
final_results.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
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
import argparse
import numpy as np
import pyquaternion as pyq
from jaxlie import SO3
import os
import copy
import json
from bop_pose_error import adi as bop_adi
from bop_pose_error import VOCap
dir_name = os.path.dirname(__file__)
class ADIEvaluator():
def __init__(self, ycbv_models_path, object_names):
self.clouds = { object_name : self.load_point_cloud(os.path.join(ycbv_models_path, 'models', object_name, 'points.xyz')) for object_name in object_names}
def load_point_cloud(self, file_name):
"""Load the xyz point cloud provided within the YCB Video Models folder."""
data = []
with open(file_name, newline = '') as file_data:
for row in file_data:
data.append([float(num_string.rstrip()) for num_string in row.rstrip().split(sep = ' ') if num_string != ''])
return np.array(data)
def adi_auc(self, object_name, reference, signal):
"""Evaluate distances and AUC for the Average Distance Indistinguishable (ADI) metric.
object_name is the object name, e.g., '002_master_chef_can'
reference is a [N, 7] matrix containing N ground truth poses each expressed as [x, y, z, axis_x, axis_y, axis_z, angle]
signal is a [N, 7] matrix containing N ground truth poses each expressed as [x, y, z, axis_x, axis_y, axis_z, angle]
Return
- the ADI distances associated to all poses
- the Area Under the Curve (AUC) of the ADI distances
"""
distances = None
sig = signal
ref = reference
dists = []
for j in range(ref.shape[0]):
gt_t = reference[j, 0 : 3]
gt_R = pyq.Quaternion(axis = reference[j, 3 : 6], angle = reference[j, 6]).rotation_matrix
estimate_t = signal[j, 0 : 3]
estimate_R = pyq.Quaternion(axis = signal[j, 3 : 6], angle = signal[j, 6]).rotation_matrix
dists.append(bop_adi(estimate_R, estimate_t, gt_R, gt_t, self.clouds[object_name]))
distances = np.array(dists)
distances_copy = copy.deepcopy(distances)
threshold = 0.02
inf_indexes = np.where(distances > threshold)[0]
distances[inf_indexes] = np.inf
sorted_distances = np.sort(distances)
n = len(sorted_distances)
accuracy = np.cumsum(np.ones((n, ), np.float32)) / n
return distances_copy, VOCap(sorted_distances, accuracy) * 100.0
def calculate_auc(object_string, poses, num_poses, ycb_video_models_path, ground_truth_position, object_names):
evaluator = ADIEvaluator(ycb_video_models_path, object_names)
gt = np.concatenate((ground_truth_position, np.array([1, 0, 0, 0])))
gt = np.repeat(np.expand_dims(gt, 0), num_poses, axis=0)
dists, auc = evaluator.adi_auc(object_string, gt, poses)
return auc
def calculate_auc_rot(object_string, poses, num_poses, ycb_video_models_path, ground_truth_position, object_names, bool_val = False):
evaluator = ADIEvaluator(ycb_video_models_path, object_names)
poses[:,:3] = ground_truth_position
gt = np.concatenate((ground_truth_position, np.array([1, 0, 0, 0])))
gt = np.repeat(np.expand_dims(gt, 0), num_poses, axis=0)
dists, auc = evaluator.adi_auc(object_string, gt, poses)
return auc
def create_table_all_objects(dict_objects, results_directory, list_keys):
list_objects_tex = []
for obj in list_keys:
list_words = obj.rsplit('_')
create_string = [x + r'{\_}'for x in list_words[:-1]]
create_string.append(list_words[-1])
create_string.append(' & ')
list_objects_tex.append(''.join(create_string))
with open(results_directory + "/table_all_objects.tex", 'w') as tf:
tf.write(r'\begin{table*}')
tf.write('\n')
tf.write(r'\tiny')
tf.write('\n')
tf.write(r'\scriptsize')
tf.write('\n')
tf.write(r'\centering')
tf.write('\n')
tf.write(r'\caption{Table 1: Comparison between the baseline and the proposed method executed on simulated data using 10 objects.}')
tf.write('\n')
tf.write(r'\begin{tabular}{| l | c | c | c | c | c | c | c | c | c | c |}')
tf.write('\n')
tf.write(r'\hline')
tf.write('\n')
tf.write(r'metric & \multicolumn {2}{c|}{Positional error (cm))} & \multicolumn {2}{c|}{ADI-AUC (\%)} & \multicolumn {2}{c|}{ADI-R-AUC (\%)} & \multicolumn{2}{c|}{Contacts}\\')
tf.write('\n')
tf.write(r'\hline')
tf.write('\n')
tf.write(r'Method & Baseline & Ours & Baseline & Ours & Baseline & Ours & Baseline & Ours \\')
tf.write('\n')
tf.write(r'\hline \\')
tf.write('\n')
tf.write(r'Evaluated poses & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 & \#1, \#1-5 \\')
tf.write('\n')
tf.write(r'\hline')
tf.write('\n')
for obj in range(len(list_objects_tex)):
tf.write(list_objects_tex[obj]
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][1])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][4])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][1])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][4])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][12])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][13])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][12])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][13])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][14])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][15])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][14])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][15])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][9])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Baseline']["Mean"][10])) + ' & '
+ str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][9])) + ', ' + str('%.2f'%(dict_objects[list_keys[obj]][4]['Sensor']["Mean"][10])))
tf.write(r'\\')
tf.write('\n')
tf.write(r'\hline')
tf.write('\n')
tf.write(r'Mean & '
+ str('%.2f'%(dict_objects["total_objects"]['Baseline']["Mean"][1])) + ', ' + str('%.2f'%(dict_objects["total_objects"]['Baseline']["Mean"][4])) + ' & '
+ str('%.2f'%(dict_objects["total_objects"]['Sensor']["Mean"][1])) + ', ' + str('%.2f'%(dict_objects["total_objects"]['Sensor']["Mean"][4])) + ' & '
+ str('%.2f'%(dict_objects['total_objects']['Baseline']["Mean"][12])) + ', ' + str('%.2f'%(dict_objects['total_objects']['Baseline']["Mean"][13])) + ' & '
+ str('%.2f'%(dict_objects['total_objects']['Sensor']["Mean"][12])) + ', ' + str('%.2f'%(dict_objects['total_objects']['Sensor']["Mean"][13])) + ' & '
+ str('%.2f'%(dict_objects['total_objects']['Baseline']["Mean"][14])) + ', ' + str('%.2f'%(dict_objects['total_objects']['Baseline']["Mean"][15])) + ' & '
+ str('%.2f'%(dict_objects['total_objects']['Sensor']["Mean"][14])) + ', ' + str('%.2f'%(dict_objects['total_objects']['Sensor']["Mean"][15])) + ' & '
+ str('%.2f'%(dict_objects["total_objects"]['Baseline']["Mean"][9])) + ', ' + str('%.2f'%(dict_objects["total_objects"]['Baseline']["Mean"][10])) + ' & '
+ str('%.2f'%(dict_objects["total_objects"]['Sensor']["Mean"][9])) + ', ' + str('%.2f'%(dict_objects["total_objects"]['Sensor']["Mean"][10])))
tf.write(r'\\')
tf.write('\n')
tf.write(r'\hline')
tf.write('\n')
tf.write(r'\end{tabular}')
tf.write('\n')
tf.write(r'\end{table*}')
tf.write('\n')
tf.write(r'\end{document}')
def main():
parser = argparse.ArgumentParser(description='')
parser.add_argument('--results_directory', dest='results_directory', help='path to the results directory',
type=str, required=True)
parser.add_argument('--ycb_directory', dest='ycb_directory', help='path to the ycb directory',
type=str, required=True)
parser.add_argument('--json_directory', dest='json_directory', help='path to the json directory',
type=str, required=True)
parser.add_argument('-l','--list', dest='gt',action='append')
args = parser.parse_args()
average_error = []
average_rotation_error = []
average_position_error = []
ground_truth_position = np.array([float(args.gt[0]),
float(args.gt[1]),
float(args.gt[2])])
with open(args.json_directory + 'contacts.json') as f:
contacts_dict = json.load(f)
with open(args.json_directory + "objects.json") as fo:
list_objects = json.load(fo)
average_error = np.array(average_error)
average_rotation_error = np.array(average_rotation_error)
average_position_error = np.array(average_position_error)
error_dict = {"Best": [], "Best 5": [[], [], []], "Best 10": [[], [], []], "Mean": [], "Poses":[]}
sensor_vs_baseline = {"Sensor": copy.deepcopy(error_dict), "Baseline": copy.deepcopy(error_dict)}
dict_objects = {}
for h in list_objects:
dict_objects[h] = [copy.deepcopy(sensor_vs_baseline), copy.deepcopy(sensor_vs_baseline), copy.deepcopy(sensor_vs_baseline), copy.deepcopy(sensor_vs_baseline), copy.deepcopy(sensor_vs_baseline)]
dict_objects["total_objects"] = copy.deepcopy(sensor_vs_baseline)
method_list_path = ["results_sensor/", "results_baseline/"]
method_list = ["Sensor", "Baseline"]
triplet_list_path = ["triplet_1/","triplet_2/", "triplet_3/", "triplet_4/"]
for key, _ in dict_objects.items():
if key == 'total_objects':
break
for triplet in range(len(triplet_list_path)):
for method in range(len(method_list)):
arrays = np.load(args.results_directory + "/" + key + "/" + triplet_list_path[triplet] + method_list_path[method] + 'final_results.npz')
all_rotations = arrays['total_poses']
sum_errors = arrays['sum_errors']
rotation_errors = arrays['pose_errors']
position_errors = arrays['position_errors']
ordered_poses = arrays['ordered_poses']
for j in range(10):
index_best = int(ordered_poses[j])
if j == 0:
dict_objects[key][triplet][method_list[method]]["Best"].append(sum_errors[index_best])
dict_objects[key][triplet][method_list[method]]["Best"].append(position_errors[index_best] * 100)
dict_objects[key][triplet][method_list[method]]["Best"].append(rotation_errors[index_best])
if j < 5:
dict_objects[key][triplet][method_list[method]]["Best 5"][0].append(sum_errors[index_best])
dict_objects[key][triplet][method_list[method]]["Best 5"][1].append(position_errors[index_best] * 100)
dict_objects[key][triplet][method_list[method]]["Best 5"][2].append(rotation_errors[index_best])
angle = np.linalg.norm(all_rotations[index_best, 3:])
axis_angle = np.array([all_rotations[index_best, 0], all_rotations[index_best, 1], all_rotations[index_best, 2], all_rotations[index_best, 3]/angle, all_rotations[index_best, 4]/angle, all_rotations[index_best, 5]/angle, angle])
dict_objects[key][triplet][method_list[method]]["Poses"].append(axis_angle)
contacts_dict[key][triplet][method_list[method]][j]= int(contacts_dict[key][triplet][method_list[method]][j]/3) * 100
dict_objects[key][triplet][method_list[method]]["Best 10"][0].append(sum_errors[index_best])
dict_objects[key][triplet][method_list[method]]["Best 10"][1].append(position_errors[index_best] * 100)
dict_objects[key][triplet][method_list[method]]["Best 10"][2].append(rotation_errors[index_best])
# Average of the 5 and 10 poses for the same triplet and contacts
dict_objects[key][triplet][method_list[method]]["Mean"].append(dict_objects[key][triplet][method_list[method]]["Best"][0])
dict_objects[key][triplet][method_list[method]]["Mean"].append(dict_objects[key][triplet][method_list[method]]["Best"][1])
dict_objects[key][triplet][method_list[method]]["Mean"].append(dict_objects[key][triplet][method_list[method]]["Best"][2])
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 5"][0])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 5"][1])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 5"][2])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 10"][0])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 10"][1])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(dict_objects[key][triplet][method_list[method]]["Best 10"][2])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(contacts_dict[key][triplet][method_list[method]][0])
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(contacts_dict[key][triplet][method_list[method]][:5])))
dict_objects[key][triplet][method_list[method]]["Mean"].append(np.mean(np.array(contacts_dict[key][triplet][method_list[method]])))
# Average of the best, 5 and 10 poses for all the triplets
for k in range(12):
dict_objects[key][4]['Sensor']["Mean"].append(np.mean(np.array([dict_objects[key][0]['Sensor']["Mean"][k],
dict_objects[key][1]['Sensor']["Mean"][k],
dict_objects[key][2]['Sensor']["Mean"][k],
dict_objects[key][3]['Sensor']["Mean"][k]])))
dict_objects[key][4]['Baseline']["Mean"].append(np.mean(np.array([dict_objects[key][0]['Baseline']["Mean"][k],
dict_objects[key][1]['Baseline']["Mean"][k],
dict_objects[key][2]['Baseline']["Mean"][k],
dict_objects[key][3]['Baseline']["Mean"][k]])))
dict_objects[key][4]['Sensor']["Mean"].append(calculate_auc(key, np.concatenate((np.array([dict_objects[key][0]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][1]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][2]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][3]['Sensor']["Poses"][0]])), axis=0), 4, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Baseline']["Mean"].append(calculate_auc(key, np.concatenate((np.array([dict_objects[key][0]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][1]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][2]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][3]['Baseline']["Poses"][0]])), axis=0), 4, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Sensor']["Mean"].append(calculate_auc(key, np.concatenate((np.array(dict_objects[key][0]['Sensor']["Poses"]),
np.array(dict_objects[key][1]['Sensor']["Poses"]),
np.array(dict_objects[key][2]['Sensor']["Poses"]),
np.array(dict_objects[key][3]['Sensor']["Poses"])), axis=0), 20, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Baseline']["Mean"].append(calculate_auc(key, np.concatenate((np.array(dict_objects[key][0]['Baseline']["Poses"]),
np.array(dict_objects[key][1]['Baseline']["Poses"]),
np.array(dict_objects[key][2]['Baseline']["Poses"]),
np.array(dict_objects[key][3]['Baseline']["Poses"])), axis=0), 20, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Sensor']["Mean"].append(calculate_auc_rot(key, np.concatenate((np.array([dict_objects[key][0]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][1]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][2]['Sensor']["Poses"][0]]),
np.array([dict_objects[key][3]['Sensor']["Poses"][0]])), axis=0), 4, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Baseline']["Mean"].append(calculate_auc_rot(key, np.concatenate((np.array([dict_objects[key][0]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][1]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][2]['Baseline']["Poses"][0]]),
np.array([dict_objects[key][3]['Baseline']["Poses"][0]])), axis=0), 4, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Sensor']["Mean"].append(calculate_auc_rot(key, np.concatenate((np.array(dict_objects[key][0]['Sensor']["Poses"]),
np.array(dict_objects[key][1]['Sensor']["Poses"]),
np.array(dict_objects[key][2]['Sensor']["Poses"]),
np.array(dict_objects[key][3]['Sensor']["Poses"])), axis=0), 20, args.ycb_directory, ground_truth_position, list_objects))
dict_objects[key][4]['Baseline']["Mean"].append(calculate_auc_rot(key, np.concatenate((np.array(dict_objects[key][0]['Baseline']["Poses"]),
np.array(dict_objects[key][1]['Baseline']["Poses"]),
np.array(dict_objects[key][2]['Baseline']["Poses"]),
np.array(dict_objects[key][3]['Baseline']["Poses"])), axis=0), 20, args.ycb_directory, ground_truth_position, list_objects))
for k in range(16):
arr_sensor = np.empty((0,1))
arr_baseline = np.empty((0,1))
for h in list_objects:
arr_sensor = np.append(arr_sensor, np.array([[dict_objects[h][4]['Sensor']["Mean"][k]]]), 0)
arr_baseline = np.append(arr_baseline, np.array([[dict_objects[h][4]['Baseline']["Mean"][k]]]), 0)
dict_objects["total_objects"]["Sensor"]["Mean"].append(np.mean(arr_sensor))
dict_objects["total_objects"]["Baseline"]["Mean"].append(np.mean(arr_baseline))
create_table_all_objects(dict_objects, args.results_directory, list_objects)
if __name__ == '__main__':
main()