-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
234 lines (188 loc) · 8.29 KB
/
app.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
from PIL import Image
import io
import pandas as pd
import numpy as np
import cv2
from typing import Optional
from ultralytics import YOLO
from ultralytics.utils.plotting import Annotator, colors
# Initialize the models
model_sample_detect = YOLO("./models/sample_model/best.pt")
def get_image_from_bytes(binary_image: bytes) -> Image:
"""Convert image from bytes to PIL RGB format
**Args:**
- **binary_image (bytes):** The binary representation of the image
**Returns:**
- **PIL.Image:** The image in PIL RGB format
"""
input_image = Image.open(io.BytesIO(binary_image)).convert("RGB")
return input_image
def get_bytes_from_image(image: Image) -> bytes:
"""
Convert PIL image to Bytes
Args:
image (Image): A PIL image instance
Returns:
bytes : BytesIO object that contains the image in JPEG format with quality 85
"""
return_image = io.BytesIO()
image.save(return_image, format='JPEG', quality=85) # save the image in JPEG format with quality 85
return_image.seek(0) # set the pointer to the beginning of the file
return return_image
def transform_predict_to_df(results: list, labeles_dict: dict) -> pd.DataFrame:
"""
Transform predict from yolov8 (torch.Tensor) to pandas DataFrame.
Args:
results (list): A list containing the predict output from yolov8 in the form of a torch.Tensor.
labeles_dict (dict): A dictionary containing the labels names, where the keys are the class ids and the values are the label names.
Returns:
predict_bbox (pd.DataFrame): A DataFrame containing the bounding box coordinates, confidence scores and class labels.
"""
# Transform the Tensor to numpy array
predict_bbox = pd.DataFrame(results[0].to("cpu").numpy().boxes.xyxy, columns=['xmin', 'ymin', 'xmax','ymax'])
# Add the confidence of the prediction to the DataFrame
predict_bbox['confidence'] = results[0].to("cpu").numpy().boxes.conf
# Add the class of the prediction to the DataFrame
predict_bbox['class'] = (results[0].to("cpu").numpy().boxes.cls).astype(int)
# Replace the class number with the class name from the labeles_dict
predict_bbox['name'] = predict_bbox["class"].replace(labeles_dict)
return predict_bbox
def get_model_predict(model: YOLO, input_image: Image, save: bool = False, image_size: int = 1248, conf: float = 0.5, augment: bool = False) -> pd.DataFrame:
"""
Get the predictions of a model on an input image.
Args:
model (YOLO): The trained YOLO model.
input_image (Image): The image on which the model will make predictions.
save (bool, optional): Whether to save the image with the predictions. Defaults to False.
image_size (int, optional): The size of the image the model will receive. Defaults to 1248.
conf (float, optional): The confidence threshold for the predictions. Defaults to 0.5.
augment (bool, optional): Whether to apply data augmentation on the input image. Defaults to False.
Returns:
pd.DataFrame: A DataFrame containing the predictions.
"""
# Make predictions
predictions = model.predict(
imgsz=image_size,
source=input_image,
conf=conf,
save=save,
augment=augment,
flipud= 0.0,
fliplr= 0.0,
mosaic = 0.0,
)
# Transform predictions to pandas dataframe
predictions = transform_predict_to_df(predictions, model.model.names)
return predictions
def get_model_segment(model: YOLO, input_image: Image, save: bool = False, image_size: int = 1248, conf: float = 0.25, augment: bool = False) -> pd.DataFrame:
"""
Get the predictions of a model on an input image.
Args:
model (YOLO): The trained YOLO model.
input_image (Image): The image on which the model will make predictions.
save (bool, optional): Whether to save the image with the predictions. Defaults to False.
image_size (int, optional): The size of the image the model will receive. Defaults to 1248.
conf (float, optional): The confidence threshold for the predictions. Defaults to 0.25.
augment (bool, optional): Whether to apply data augmentation on the input image. Defaults to False.
Returns:
pd.DataFrame: A DataFrame containing the predictions.
"""
# Make predictions
predictions = model.predict(
imgsz=image_size,
source=input_image,
conf=conf,
save=save,
augment=augment,
flipud= 0.0,
fliplr= 0.0,
mosaic = 0.0,
)
# Transform predictions to pandas dataframe
predictions = transform_predict_to_df(predictions, model.model.names)
return predictions
################################# BBOX Func #####################################
def add_bboxs_on_img(image: Image, predict: pd.DataFrame()) -> Image:
"""
add a bounding box on the image
Args:
image (Image): input image
predict (pd.DataFrame): predict from model
Returns:
Image: image whis bboxs
"""
# Create an annotator object
annotator = Annotator(np.array(image))
# sort predict by xmin value
predict = predict.sort_values(by=['xmin'], ascending=True)
# iterate over the rows of predict dataframe
for i, row in predict.iterrows():
# create the text to be displayed on image
text = f"{row['name']}: {int(row['confidence']*100)}%"
# get the bounding box coordinates
bbox = [row['xmin'], row['ymin'], row['xmax'], row['ymax']]
# add the bounding box and text on the image
annotator.box_label(bbox, text, color=colors(row['class'], True))
# convert the annotated image to PIL image
return Image.fromarray(annotator.result())
################################# Models #####################################
def detect_sample_model(input_image: Image) -> pd.DataFrame:
"""
Predict from sample_model.
Base on YoloV8
Args:
input_image (Image): The input image.
Returns:
pd.DataFrame: DataFrame containing the object location.
"""
predict = get_model_predict(
model=model_sample_detect,
input_image=input_image,
save=False,
image_size=768,
augment=False,
conf=0.5,
)
return predict
def censor_objects(image: Image, predictions: pd.DataFrame, method: str) -> Image:
"""
Censor detected objects in an image using OpenCV.
Args:
image (Image): The original PIL image.
predictions (pd.DataFrame): DataFrame containing detection results with bounding boxes.
method (str): Method of censorship ('blur' for Gaussian blur or 'mask' for a solid color mask). Default is 'blur'.
Returns:
Image: The censored image as a PIL image.
"""
# Convert PIL image to OpenCV format (PIL uses RGB, OpenCV uses BGR)
open_cv_image = np.array(image.convert('RGB'))[:, :, ::-1]
for index, row in predictions.iterrows():
# Extract bounding box coordinates
x1, y1, x2, y2 = int(row['xmin']), int(row['ymin']), int(row['xmax']), int(row['ymax'])
if method == 'blur':
# Apply Gaussian blur to the specified area (bounding box)
open_cv_image[y1:y2, x1:x2] = cv2.GaussianBlur(open_cv_image[y1:y2, x1:x2], (0, 0), 10)
elif method == 'mask':
# Apply a solid color mask (black) to the specified area
open_cv_image[y1:y2, x1:x2] = 0
# Convert back to PIL image from OpenCV format
censored_image = Image.fromarray(open_cv_image[:, :, ::-1])
return censored_image
def segment_sample_model(input_image: Image) -> pd.DataFrame:
"""
Predict from sample_model.
Base on YoloV8
Args:
input_image (Image): The input image.
Returns:
pd.Dataframe: Dataframe containing the object location and segmentation.
"""
predict = get_model_segment(
model=model_sample_segment,
input_image=input_image,
save=True,
image_size=640,
augment=False,
conf=0.25,
)
return predict