-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathtest_cd.py
158 lines (139 loc) · 7.46 KB
/
test_cd.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
import torch
from torch import autocast
from torchvision import transforms
from torch.utils.data import Dataset, DataLoader
import json
import pandas as pd
from PIL import Image
import numpy as np
from chamfer_distance import ChamferDistance
import open3d as o3d
import sys
import os
sys.path.append("/home/ubuntu/data/abo")
from common.io_assets.eames.load_glb import eames_load_asset
SAVE_DIR = "/home/ubuntu/workspace/zero123/3drec/experiments/finetune_1000steps_sofa_zero123_renderings"
class ABODataset(Dataset):
def __init__(self, threshold=16):
self.threshold = threshold
metadata_csv = pd.read_csv('/home/ubuntu/data/abo/3dmodels/metadata/3dmodels.csv')
self.model_index_to_path = dict()
for i in range(len(metadata_csv)):
self.model_index_to_path[metadata_csv.iloc[i]['3dmodel_id']] = metadata_csv.iloc[i]['path']
self.meta = []
split = pd.read_csv('/home/ubuntu/data/abo/render/train_test_split.csv')
split = split[split['SPLIT'] == 'TEST']['MODEL']
for model_name in split:
# if not os.path.exists(f"{SAVE_DIR}/scene-{model_name}-index-0_scale-100.0_train-view-True_view-weight-10000_depth-smooth-wt-10000.0_near-view-wt-10000.0/model.obj"):
# continue
with open(f'/home/ubuntu/data/abo/angle/{model_name}.json', "r") as f:
angle_json = json.load(f)
with open(f'/home/ubuntu/data/abo/inference_index/{model_name}.json', "r") as f:
inference_json = json.load(f)
for env_index in inference_json:
# for inference_id, target_id in zip(inference_json[env_index]["view_inference"], inference_json[env_index]["view_target"]):
# inference_id = str(inference_id)
# target_id = str(target_id)
# meta_dict = {
# "model_name": model_name,
# "inference_index": inference_id,
# "target_index": target_id,
# "env_index": env_index,
# "inference_image_path": f"/home/ubuntu/data/abo/render/{model_name}/render/{env_index}/render_{inference_id}.jpg",
# "inference_mask_path": f"/home/ubuntu/data/abo/render/{model_name}/segmentation/segmentation_{inference_id}.jpg",
# "target_image_path": f"/home/ubuntu/data/abo/render/{model_name}/render/{env_index}/render_{target_id}.jpg",
# "target_mask_path": f"/home/ubuntu/data/abo/render/{model_name}/segmentation/segmentation_{target_id}.jpg",
# "gen_image_path": f"{SAVE_DIR}/{model_name}/{env_index}/{target_id}.jpg",
# "inference_angle": angle_json[inference_id]["angle"],
# "target_angle": angle_json[target_id]["angle"],
# "azimuth_offset": (angle_json[target_id]["angle"][0]-angle_json[inference_id]["angle"][0])%(2*np.pi),
# "elevation_offset": (angle_json[target_id]["angle"][1]-angle_json[inference_id]["angle"][1]),
# "distance_offset": angle_json[target_id]["angle"][2]-angle_json[inference_id]["angle"][2]
# }
# self.meta.append(meta_dict)
inference_id = str(22)
if os.path.exists(f'/home/ubuntu/workspace/zero123/3drec/data/sofa/{model_name}/{env_index}_{inference_id}.jpg'):
meta_dict = {
"model_name": model_name,
"inference_index": inference_id,
"env_index": env_index,
"inference_image_path": f"/home/ubuntu/data/abo/render/{model_name}/render/{env_index}/render_{inference_id}.jpg",
"inference_mask_path": f"/home/ubuntu/data/abo/render/{model_name}/segmentation/segmentation_{inference_id}.jpg",
"inference_angle": angle_json[inference_id]["angle"],
}
self.meta.append(meta_dict)
def __len__(self):
return len(self.meta)
def __getitem__(self, idx):
meta_info = self.meta[idx]
model_name, env_index = meta_info['model_name'], meta_info['env_index']
model_path = self.model_index_to_path[model_name]
target_path = f"/home/ubuntu/data/abo/3dmodels/original/{model_path}"
json_path = f"/home/ubuntu/data/abo/render/{model_name}/metadata.json"
target_points = self.load_normalized_glb(target_path, json_path=json_path)
gen_path = f"{SAVE_DIR}/scene-{model_name}-index-0_scale-100.0_train-view-True_view-weight-10000_depth-smooth-wt-10000.0_near-view-wt-10000.0/model_m{self.threshold}.obj"
generation_points = self.load_normalized_obj(gen_path)
target_points = np.asarray(target_points.points)
generation_points = np.asarray(generation_points.points)
return torch.tensor(target_points), torch.tensor(generation_points)#, model_name, env_index, target_index
def load_normalized_glb(self, glb_path, json_path):
# with open(json_path, "r") as f:
# render_json = json.load(f)
open3d_obj = eames_load_asset(glb_path)
# exhange_axis = np.array([[-1,0,0],[0,0,-1],[0,1,0]])
verts = np.asarray(open3d_obj.vertices)
# verts = verts@exhange_axis
vmin = verts.min(axis=0)
vmax = verts.max(axis=0)
vcen = (vmin+vmax)/2
obj_size = np.abs(verts - vcen).max()
verts = verts - vcen.reshape(1,3)
verts = verts/obj_size
open3d_obj.vertices = o3d.utility.Vector3dVector(verts)
points = open3d_obj.sample_points_uniformly(10000)
return points
def load_normalized_obj(self, obj_path):
open3d_obj = o3d.io.read_triangle_mesh(obj_path)
verts = np.asarray(open3d_obj.vertices)
vmin = verts.min(axis=0)
vmax = verts.max(axis=0)
vcen = (vmin+vmax)/2
obj_size = np.abs(verts - vcen).max()
verts = verts - vcen.reshape(1,3)
verts = verts/obj_size
open3d_obj.vertices = o3d.utility.Vector3dVector(verts)
points = open3d_obj.sample_points_uniformly(10000)
return points
if __name__ == '__main__':
chamfer_dist = ChamferDistance()
cd_list = {'20':[],
'18':[],
'16':[],
'14':[],
'12':[],
'10':[],
'8':[],
'6':[],
'4':[],
'2':[]
}
for threshold in cd_list:
print("Calculate threshold", threshold)
dataset = ABODataset(threshold=int(threshold))
dataloader = DataLoader(dataset, batch_size=2, shuffle=False)
for data in dataloader:
real_points, gen_points = data#, model_name, env_index, target_index = data
real_points = real_points.cuda()
gen_points = gen_points.cuda()
dist1, dist2, idx1, idx2 = chamfer_dist(real_points, gen_points)
loss = (torch.mean(dist1, dim=1)) + (torch.mean(dist2, dim=1))
cd_list[threshold].extend(loss.cpu().tolist())
data_num = len(cd_list[threshold])
cd = np.mean(cd_list[threshold])
print(f"Size: {data_num}")
print(f"CD: {cd}")
with open(f"{SAVE_DIR}/test_all_threshold.txt", "a") as f:
f.write(f"Threshold: {threshold}\n")
f.write(f"Size: {data_num}\n")
f.write(f"CD: {cd}\n")
f.write(f"************\n")