-
Notifications
You must be signed in to change notification settings - Fork 0
/
write_csv_new.py
196 lines (155 loc) · 7.67 KB
/
write_csv_new.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
import sys
from models.pose_model import PoseModel
from dataset.transform import ImageTransform
from dataset.draw import PredictionVisualizer
from utils.utils import get_option_path
from PIL import Image
import cv2
import os
from utils.utils import get_corresponding_cfg
import torch
import csv
import json
posenet = PoseModel()
class ImageVisualizer:
out_h, out_w, in_h, in_w = 64, 64, 256, 256
def __init__(self, model_cfg, model_path, data_cfg=None, show=True, device="cuda", conf=0.05):
self.show = show
self.device = device
option_file = get_option_path(model_path)
self.transform = ImageTransform()
if os.path.exists(option_file):
option = torch.load(option_file)
self.out_h, self.out_w, self.in_h, self.in_w = \
option.output_height, option.output_width, option.input_height, option.input_width
else:
if data_cfg:
self.transform.init_with_cfg(data_cfg)
self.out_h, self.out_w, self.in_h, self.in_w = \
self.transform.output_height, self.transform.output_width, self.transform.input_height,self.transform.input_width
else:
pass
posenet.build(model_cfg)
self.model = posenet.model
self.kps = posenet.kps
self.model.eval()
if isinstance(conf, float):
self.conf = torch.tensor([conf for _ in range(self.kps)])
else:
self.conf = torch.tensor([float(i) for i in conf.split(",")])
self.conf = conf
posenet.load(model_path)
self.PV = PredictionVisualizer(posenet.kps, 1, self.out_h, self.out_w, self.in_h, self.in_w, max_img=1, column=1)
def img_splitcate(self, json_path, scale_factor, folder_path, cate_output):
with open(json_path, "r") as f:
data = json.load(f)
image_files = os.listdir(folder_path)
os.makedirs(cate_output, exist_ok=True)
category_count = {}
for image_data in data["images"]:
image_id = image_data["id"]
file_name = image_data["file_name"]
if file_name in image_files:
image_path = os.path.join(folder_path, file_name)
annotation = next((ann for ann in data["annotations"] if ann["image_id"] == image_id), None)
if annotation:
category_id = annotation["category_id"]
bbox = annotation["bbox"]
category_name = next((cat["name"] for cat in data["categories"] if cat["id"] == category_id), None)
if category_name:
category_folder = os.path.join(cate_output, category_name)
os.makedirs(category_folder, exist_ok=True)
if category_name in category_count:
category_count[category_name] += 1
else:
category_count[category_name] = 1
new_file_name = f"{os.path.splitext(file_name)[0]}_{category_name}_{category_count[category_name]}.jpg"
image = Image.open(image_path)
scale_factor /= 2
width_scale = bbox[2] * scale_factor
height_scale = bbox[3] * scale_factor
left = max(0, bbox[0] - width_scale)
top = max(0, bbox[1] - height_scale)
right = min(image.width, bbox[0] + bbox[2] + width_scale)
bottom = min(image.height, bbox[1] + bbox[3] + height_scale)
if left < right and top < bottom:
cropped_image = image.crop((left, top, right, bottom))
cropped_image.save(os.path.join(category_folder, new_file_name))
else:
print(f"Invalid coordinates for cropping: {bbox}")
else:
print(f"Cannot find category name for category_id: {category_id}")
else:
print(f"No annotation found for image_id: {image_id}")
else:
print(f"Image file not found: {file_name}")
def visualize(self, img_path, label_file, csv_path, folder_path):
with torch.no_grad():
img = cv2.imread(img_path)
inp, padded_size = self.transform.process_single_img(img_path, self.out_h, self.out_w, self.in_h, self.in_w)
img_meta = {
"name": img_path,
"enlarged_box": [0, 0, img.shape[1], img.shape[0]],
"padded_size": padded_size
}
if self.device != "cpu":
inp = inp.cuda()
out = self.model(inp.unsqueeze(dim=0))
location, img_h, img_w = self.PV.draw_kps_csv(out, img_meta, self.conf)
max_value = self.PV.getPrediction(out)[1]
if_exist = [(v>c).tolist() for c, v in zip(self.conf, max_value.squeeze())]
float_numbers = [float(i) for i in location.flatten().tolist()]
modified_array = []
for index, num in enumerate(float_numbers):
if if_exist[int(index/2)] is True:
if index % 2 == 0:
modified_array.append(num / img_w)
else:
modified_array.append(num / img_h)
else:
modified_array.append(-1)
with open(label_file, 'r') as label:
cate_array = label.readlines()
folder_cate = os.path.basename(folder_path)
for idx, cate in enumerate(cate_array):
if cate[:-1] == folder_cate:
modified_array.extend([f"{idx}", f"{cate[:-1]}", filename])
# print(modified_array)
with open(csv_path, 'a', newline='') as file:
writer = csv.writer(file)
writer.writerow(modified_array)
if __name__ == '__main__':
# Pose model path
model_path = "/home/hkuit164/Desktop/xjl/1025+1103+1121/alphapose/latest.pth"
folder_path = "/media/hkuit164/Backup/xjl/ML_data_process/Label_Studio/result/val/images"
json_path = "/media/hkuit164/Backup/xjl/ML_data_process/Label_Studio/result/val/result.json"
label_path = "/media/hkuit164/Backup/xjl/ML_data_process/ML/csv/label"
# output
csv_path = "/media/hkuit164/Backup/xjl/ML_data_process/ML/csv/test.csv"
cate_output = "/media/hkuit164/Backup/xjl/ML_data_process/ML/csv/test"
conf = 0.05
model_cfg = ""
data_cfg = ""
option_path = ""
if not model_path or not data_cfg:
model_cfg, data_cfg, option_path = get_corresponding_cfg(model_path, check_exist=["data", "model"])
if os.path.exists(option_path):
info = torch.load(option_path)
if "thresh" in info:
conf = info.thresh
IV = ImageVisualizer(model_cfg, model_path, data_cfg, conf=conf)
with open(data_cfg, "r") as load_f:
load_dict = json.load(load_f)
scale_factor = load_dict["scale"]
IV.img_splitcate(json_path, scale_factor, folder_path, cate_output)
for img_folder_name in os.listdir(cate_output):
img_folder_path = os.path.join(cate_output, img_folder_name)
if os.path.isdir(img_folder_path):
for idx, filename in enumerate(os.listdir(img_folder_path)):
if filename.endswith(".jpg"):
img_path = os.path.join(img_folder_path, filename)
try:
IV.visualize(img_path, label_path, csv_path, img_folder_path)
except:
print(idx)
sys.exit(1)