인공지능 – Lecture Note

YOLO v5를 사용하여 포트홀 찾기

참조 : https://www.youtube.com/watch?v=fdWx3QV5n44&t=4231s YOLO v5 모델(PyTorch) https://github.com/ultralytics/yolov5 https://www.ultralytics.com 테스트 환경 : Google Colab 모델 다운로드

%cd /content
!git clone https://github.com/ultralytics/yolov5
%cd yolov5
%pip install -qr requirements.txt

import torch
from IPython.display import Image, clear_output

%cd /content

!git clone https://github.com/ultralytics/yolov5

%cd yolov5

%pip install -qr requirements.txt

import torch

from IPython.display import Image, clear_output

추론(Inference) : detect.py: YOLO v5 모델을 이용한 추론 !ls data/images Image(filename=’data/images/bus.jpg’, width=600) Image(filename=’data/images/zidane.jpg’, width=600) !python detect.py –weights yolov5s.pt –img 640 –conf 0.25 –source data/images/ Image(filename=’runs/detect/exp/bus.jpg’, width=600) Image(filename=’runs/detect/exp/zidane.jpg’, width=600) 결과 Pothole 탐지 모델 데이터셋 다운로드 포트홀 데이터셋: https://public.roboflow.com/object-detection/pothole 회원가입해야 함 데이터 셋 다운로드

%mkdir /content/yolov5/pothole
%cd /content/yolov5/pothole
!wget https://public.roboflow.com/ds/ZnuYZWmewn?key=z6G9lWGoPf -O roboflow.zip
!unzip roboflow.zip
!rm roboflow.zip

%mkdir /content/yolov5/pothole

%cd /content/yolov5/pothole

!wget https://public.roboflow.com/ds/ZnuYZWmewn?key=z6G9lWGoPf -O roboflow.zip

!unzip roboflow.zip

!rm roboflow.zip

이미지 목록 생성 : 465,… Continue Reading YOLO v5를 사용하여 포트홀 찾기

폐암 수술 환자의 생존율 예측하기

사용 데이터셋: ThoraricSurgery.csv 소스 코드 1

from keras.models import Sequential
from keras.layers import Dense

import numpy as np
import tensorflow as tf

seed = 3
np.random.seed(seed)      # 항상 똑같은 난수가 나오도록 설정
tf.random.set_seed(seed)  # 항상 똑같은 난수가 나오도록 설정

Data_set = np.loadtxt("ThoraricSurgery.csv", delimiter=",")

X = Data_set[:, 0:17]
Y = Data_set[:, 17]

model = Sequential()
model.add(Dense(30, input_dim=17, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])
model.fit(X, Y, epochs=200, batch_size=10)

score = model.evaluate(X, Y)

print("\n Accuracy: %.4f \n loss    : %.4f" % (score[1], score[0]))

from keras.models import Sequential

from keras.layers import Dense

import numpy as np

import tensorflow as tf

seed = 3

np.random.seed(seed) # 항상 똑같은 난수가 나오도록 설정

tf.random.set_seed(seed) # 항상 똑같은 난수가 나오도록 설정

Data_set = np.loadtxt("ThoraricSurgery.csv", delimiter=",")

X = Data_set[:, 0:17]

Y = Data_set[:, 17]

model = Sequential()

model.add(Dense(30, input_dim=17, activation='relu'))

model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])

model.fit(X, Y, epochs=200, batch_size=10)

score = model.evaluate(X, Y)

print("\n Accuracy: %.4f \n loss : %.4f" % (score[1], score[0]))

소스 코드 1 실행 결과

...
...
Epoch 199/200
47/47 [==============================] - 0s 1ms/step - loss: 0.1117 - accuracy: 0.8702
Epoch 200/200
47/47 [==============================] - 0s 2ms/step - loss: 0.1146 - accuracy: 0.8745
15/15 [==============================] - 0s 2ms/step - loss: 0.1264 - accuracy: 0.8574

 Accuracy: 0.8574
 loss    : 0.1264

...

Epoch 199/200

47/47 [==============================] - 0s 1ms/step - loss: 0.1117 - accuracy: 0.8702

Epoch 200/200

47/47 [==============================] - 0s 2ms/step - loss: 0.1146 - accuracy: 0.8745

15/15 [==============================] - 0s 2ms/step - loss: 0.1264 - accuracy: 0.8574

Accuracy: 0.8574

loss : 0.1264

설명 accuracy : 예측이 성공학 확률 loss : 예측이 실패할 확률 예측 성공률은 데이터를 분석해 데이터를 확장하거나, 딥러닝 구조를 적절하게 바꾸는 등의 노력으로 더 향상될 수 있음 소스 코드 2

from keras.models import Sequential
from keras.layers import Dense

import numpy as np
import tensorflow as tf

from sklearn.model_selection import train_test_split

seed = 3
np.random.seed(seed)      # 항상 똑같은 난수가 나오도록 설정
tf.random.set_seed(seed)  # 항상 똑같은 난수가 나오도록 설정

Data_set = np.loadtxt("ThoraricSurgery.csv", delimiter=",")

X = Data_set[:, 0:17]
Y = Data_set[:, 17]

X, x_test, Y, y_test = train_test_split(X, Y, train_size=0.8)

model = Sequential()
model.add(Dense(30, input_dim=17, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])
model.fit(X, Y, epochs=200, batch_size=10)

score = model.evaluate(x_test, y_test)

print("\n Accuracy: %.4f \n loss    : %.4f" % (score[1], score[0]))

from keras.models import Sequential

from keras.layers import Dense

import numpy as np

import tensorflow as tf

from sklearn.model_selection import train_test_split

seed = 3

np.random.seed(seed) # 항상 똑같은 난수가 나오도록 설정

tf.random.set_seed(seed) # 항상 똑같은 난수가 나오도록 설정

Data_set = np.loadtxt("ThoraricSurgery.csv", delimiter=",")

X = Data_set[:, 0:17]

Y = Data_set[:, 17]

X, x_test, Y, y_test = train_test_split(X, Y, train_size=0.8)

model = Sequential()

model.add(Dense(30, input_dim=17, activation='relu'))

model.add(Dense(1, activation='sigmoid'))

model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])

model.fit(X, Y, epochs=200, batch_size=10)

score = model.evaluate(x_test, y_test)

print("\n Accuracy: %.4f \n loss : %.4f" % (score[1], score[0]))

소스 코드 2 실행 결과

...
...
Epoch 199/200
38/38 [==============================] - 0s 1ms/step - loss: 0.1170 - accuracy: 0.8644
Epoch 200/200
38/38 [==============================] - 0s 1ms/step - loss: 0.1143 - accuracy: 0.8723
3/3 [==============================] - 0s 5ms/step - loss: 0.1222 - accuracy: 0.8617

 Accuracy: 0.8617
 loss    : 0.1222

...

Epoch 199/200

38/38 [==============================] - 0s 1ms/step - loss: 0.1170 - accuracy: 0.8644

Epoch 200/200

38/38 [==============================] - 0s 1ms/step - loss: 0.1143 - accuracy: 0.8723

3/3 [==============================] - 0s 5ms/step - loss: 0.1222 - accuracy: 0.8617

Accuracy: 0.8617

loss : 0.1222

… Continue Reading 폐암 수술 환자의 생존율 예측하기