[AI] tensorflow (3)

가위바위보 예제

 

1. day15

1. gawi_mnist

user가 무조건 이기는 코드 import tensorflow as tf import numpy as np x_train = [ [1, 0, 0], #가위 [0, 1, 0], #바위 [0, 0, 1], #보 ] y_train = [2, 0, 1] x_test = [ [1, 0, 0] #가위 ] x_train_np = np.array(x_train, dtype='f') #모델의 입맛에 맞추기 위해 실수로 바꿈 / 배열은 모델안에 들어갈 수 없어서 numpy로 바꿔줌 y_train_np = np.array(y_train, dtype='f') x_test_np = np.array(x_test, dtype='f') #모델 구성 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(3,)), tf.keras.layers.Dense(25, activation=tf.nn.relu), tf.keras.layers.Dense(25, activation=tf.nn.relu), tf.keras.layers.Dense(3, activation=tf.nn.softmax) ]) #4. 모델 컴파일 model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) #모델 훈련 model.fit(x_train, y_train, epochs=15) model.save("gawi.h5") predict = model.predict(x_test_np) idx_pred = np.argmax(predict[0]) print(predict)

x_train = [             [1, 0, 0],  #가위             [0, 1, 0],  #바위             [0, 0, 1],  #보 ] y_train = [2, 0, 1] :  x_train이 가위 [1,0,0]일때 y_train은 2로 출력 (why 2? 가위-보여야하니까 보[0 0 1]의 argmax값 2)                                나머지는 그림 참고

 

2. gawi_ai

import numpy as np import tensorflow as tf import keras mine="" com = "" result = "" #선언부 mine = input("가위/바위/보를 입력하세요.") x_test = [] if mine == "가위": x_test.append([1, 0, 0]) elif mine == "바위": x_test.append([0, 1, 0]) else: x_test.append([0, 0, 1]) x_test_np = np.array(x_test) model = keras.models.load_model("gawi.h5") predict = model.predict(x_test_np) inx_gw = np.argmax(predict[0]) if inx_gw == 0: com = "가위" elif inx_gw == 1: com = "바위" else: com = "보" if com == "가위" and mine == "가위" : result = "비김" if com == "가위" and mine == "바위" : result = "승리" if com == "가위" and mine == "보" : result = "패배" if com == "바위" and mine == "가위" : result = "패배" if com == "바위" and mine == "바위" : result = "비김" if com == "바위" and mine == "보" : result = "승리" if com == "보" and mine == "가위" : result = "승리" if com == "보" and mine == "바위" : result = "패배" if com == "보" and mine == "보" : result = "비김" print("나:",mine) print("컴:",com) print("결과:",result)

 

 

2. day16

 

1. gawi_mnist

import tensorflow as tf import numpy as np x_train = [ [0], #가위 [0.5], #바위 [1], #보 ] y_train = [2, 0, 1] #이기는 게임 gawi_ai # y_train = [0, 1, 2] x_test = [ [0] #가위 ] x_train_np = np.array(x_train, dtype='f') #모델의 입맛에 맞추기 위해 실수로 바꿈 / 배열은 모델안에 들어갈 수 없어서 numpy로 바꿔줌 y_train_np = np.array(y_train, dtype='f') x_test_np = np.array(x_test, dtype='f') #모델 구성 model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(1,)), tf.keras.layers.Dense(25, activation=tf.nn.relu), tf.keras.layers.Dense(25, activation=tf.nn.relu), tf.keras.layers.Dense(3, activation=tf.nn.softmax) ]) #4. 모델 컴파일 model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) #모델 훈련 model.fit(x_train, y_train, epochs=214) model.save("gawi.h5") predict = model.predict(x_test_np) idx_pred = np.argmax(predict[0]) print(predict)

x_train = [             [0],  #가위             [0.5],  #바위             [1],  #보 ]

2. gawi_ai

import numpy as np import tensorflow as tf import keras mine="" com = "" result = "" #선언부 mine = input("가위/바위/보를 입력하세요.") x_test = [] if mine == "가위": x_test.append([0]) elif mine == "바위": x_test.append([0.5]) else: x_test.append([1]) x_test_np = np.array(x_test) model = keras.models.load_model("gawi.h5") predict = model.predict(x_test_np) inx_gw = np.argmax(predict[0]) if inx_gw == 0: com = "가위" elif inx_gw == 1: com = "바위" else: com = "보" if com == "가위" and mine == "가위" : result = "비김" if com == "가위" and mine == "바위" : result = "승리" if com == "가위" and mine == "보" : result = "패배" if com == "바위" and mine == "가위" : result = "패배" if com == "바위" and mine == "바위" : result = "비김" if com == "바위" and mine == "보" : result = "승리" if com == "보" and mine == "가위" : result = "승리" if com == "보" and mine == "바위" : result = "패배" if com == "보" and mine == "보" : result = "비김" print("나:",mine) print("컴:",com) print("결과:",result)

 

 

3. mygui09_ai

무조건 비기는 가위바위보 import sys from PyQt5.QtWidgets import * from PyQt5 import uic import numpy as np import tensorflow as tf import keras from day16.gawi_aao import AaoGawi form_class = uic.loadUiType("mygui09.ui")[0] class MyWindow(QMainWindow, form_class): def __init__(self): super().__init__() self.ag = AaoGawi() self.ag.load() self.setupUi(self) self.pb.clicked.connect(self.myclick) self.leMine.returnPressed.connect(self.myclick) #엔터칠때 클릭이랑 같은 기능 def myclick(self): mine="" com = "" result = "" mine = self.leMine.text() # com = self.leCom.text() # result = self.leResult.text() x_test = [] if mine == "가위": x_test.append([0]) elif mine == "바위": x_test.append([0.5]) else: x_test.append([1]) x_test_np = np.array(x_test) predict = self.ag.model.predict(x_test_np) inx_gw = np.argmax(predict[0]) if inx_gw == 0: com = "가위" elif inx_gw == 1: com = "바위" else: com = "보" if com == "가위" and mine == "가위" : result = "비김" if com == "가위" and mine == "바위" : result = "승리" if com == "가위" and mine == "보" : result = "패배" if com == "바위" and mine == "가위" : result = "패배" if com == "바위" and mine == "바위" : result = "비김" if com == "바위" and mine == "보" : result = "승리" if com == "보" and mine == "가위" : result = "승리" if com == "보" and mine == "바위" : result = "패배" if com == "보" and mine == "보" : result = "비김" self.leCom.setText(com) self.leResult.setText(result) if __name__ == "__main__": app = QApplication(sys.argv) myWindow = MyWindow() myWindow.show() app.exec_()

 

4. gawi_aao

import tensorflow as tf import numpy as np import keras #전역으로 빼기 class AaoGawi: def __init__(self): self.model = None def load(self): self.model = keras.models.load_model("gawi.h5") def run_makeh5(self): x_train = [ [0], #가위 [0.5], #바위 [1], #보 ] # y_train = [2, 0, 1] #이기는 게임 gawi_ai y_train = [0, 1, 2] x_test = [ [0] #가위 ] x_train_np = np.array(x_train, dtype='f') #모델의 입맛에 맞추기 위해 실수로 바꿈 / 배열은 모델안에 들어갈 수 없어서 numpy로 바꿔줌 y_train_np = np.array(y_train, dtype='f') x_test_np = np.array(x_test, dtype='f') #모델 구성 self.model = tf.keras.models.Sequential([ tf.keras.layers.Flatten(input_shape=(1,)), tf.keras.layers.Dense(25, activation=tf.nn.relu), tf.keras.layers.Dense(25, activation=tf.nn.relu), #히든 부분을 여러번 해서 심층화된 학습을 시키려고 ? 아무튼 이런 뜻 여러번 쓸수록 정확도가 높아짐 but 속도가 느려짐 tf.keras.layers.Dense(3, activation=tf.nn.softmax) ]) #4. 모델 컴파일 self.model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy']) #모델 훈련 self.model.fit(x_train, y_train, epochs=214) self.model.save("gawi.h5") if __name__ == '__main__': ag = AaoGawi() ag.run_makeh5()