Question

我正在尝试实现一个tensorflow项目，其中用户绘制一个数字，模型可以识别它。到目前为止，我已经让用户在tkinter画布上绘制数字，保存图像，然后再次将该图像转换为numpy数组以进行模型预测。但是我一直收到以下错误： ValueError Input 0 of layer dense is incompatible with the layer: expected axis of -1 of input shape to have value 784 but received input with shape [None, 840]

这是代码：

def draw(event):
    color='black'
    x1, y1 = (event.x-1), (event.y-1)
    x2, y2 = (event.x+1), (event.y+1)
    c.create_rectangle(x1,y1,x2,y2, fill=color, width=10)
    blank.rectangle([x1-4,y1-4,x2+4,y2+4], fill=color, width=10)

def end(event):
    filename = "myNum.jpg"
    myImage.save(filename)                  #save image
    print("image saved")
    data = image.imread("myNum.jpg")        #convert image to numpy array
    predictModel(data)

def predictModel(data):
    model = tf.keras.models.load_model('numbers.model')
    prediction = model.predict(data)
    print(np.argmax(prediction[0]))

if __name__ == "__main__":
    root = Tk()
    c = Canvas(root, height=280, width=280, bg='white')
    c.pack()
    myImage = Image.new("RGB", (280, 280), 'white')
    blank = ImageDraw.Draw(myImage)
    root.bind('<Return>', end)
    c.bind('<B1-Motion>', draw)
    root.mainloop()

我正在使用此视频中的预测模型：https://www.youtube.com/watch?v=wQ8BIBpya2k&

model = keras.models.Sequential()
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(128, activation=tf.nn.relu))         #I think the error is occurring here
model.add(keras.layers.Dense(128, activation=tf.nn.relu))
model.add(keras.layers.Dense(10, activation=tf.nn.softmax))

非常感谢您的帮助

密集层的输入0与该层不兼容：输入形状的-1的预期轴的值为784，但接收到形状为[None，840]的输入

0 个答案: