为什么我的 python 代码出现值错误

Question

使用过：tensorflow、keras、python、flask

Error: ValueError: Input 0 of layer conv2d_3 is incompatible with the layer: : expected min_ndim=4, found ndim=3. Full shape received: (32, 224, 1)

神经网络结构：

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
rescaling_1 (Rescaling)      (None, 224, 224, 3)       0         
_________________________________________________________________
conv2d (Conv2D)              (None, 224, 224, 16)      448       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 112, 112, 32)      4640      
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 56, 56, 64)        18496     
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 28, 28, 64)        0         
_________________________________________________________________
flatten (Flatten)            (None, 50176)             0         
_________________________________________________________________
dense (Dense)                (None, 128)               6422656   
_________________________________________________________________
dense_1 (Dense)              (None, 2)                 258       
=================================================================
Total params: 6,446,498
Trainable params: 6,446,498
Non-trainable params: 0
_________________________________________________________________

from flask import *
import h5py
from PIL import Image, ImageFile
from io import BytesIO
from werkzeug.utils import secure_filename
import numpy as np
from matplotlib import image
import os
import uuid
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
from werkzeug.wrappers import Request, Response
import tensorflow as tf
import numpy as np
IMGSIZE=224
IMGCHANNELS=3*224*224
ALLOWEDEXTENSIONS=['png','jpg','txt','pdf','jfif','jpeg','gif']
classes=('NORMAL','PNEMONIAL')
app=Flask(__name__)
model=load_model(os.path.join(os.getcwd(),"chestXray.h5"))
def preprocess(inputs):
    inputs=np.array(inputs)
    l=inputs.size
    for i in range(l):
        j=inputs[i]
        j /=255
        inputs[i]=j
        j=0
    return inputs
#Allowed file function
def allowededFile(filename):
    return '.' in filename and filename.rsplit('.',1)[1] in ALLOWEDEXTENSIONS
#Flask ROutes
@app.route("/")
def index():
    return render_template("chest.html", Prediction=" ")
@app.route("/api/image", methods=['GET','POST'])
def api():    
    file=request.files['file']
    if file.filename=='':
        return render_template("chest.html", Prediction="You did not upload a file")
    if file and allowededFile(file.filename):
        filename=secure_filename(file.filename)
        print("***"+filename)
        x=[]
        
        ImageFile.LOAD_TRUNCATED_IMAGES=False
        img=Image.open(BytesIO(file.read()))
        img.load()
        img=img.resize((IMGSIZE,IMGSIZE),Image.ANTIALIAS)
        x=image.img_to_array(img)
        
        #np.expand_dims(x,axis=-1)
        x.reshape((1,224,224,1))     
        pred=model.predict(x)
        lst=decode_predictions(pred,top=3)
        items=[]
        for item in lst[0]:
            items.append({'name':item[1],'probability':float(item[2])})
        response={'prediction':items}
        return render_template("chest.html",Prediction="The state of this person's lungs is most likely {}".format(response))       
    else:
        return render_template("chest.html",Prediction='Invalid file extension')
    
if __name__=="__main__":
    from werkzeug.serving import run_simple
    run_simple('localhost',9000,app)

型号代码：

model=Sequential([
    layers.experimental.preprocessing.Rescaling(1./255,input_shape=(imgSize,imgSize,3)),
    layers.Conv2D(16,3,padding='same',activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(32,3,padding='same',activation='relu'),
    layers.MaxPooling2D(),
    layers.Conv2D(64,3,padding='same',activation='relu'),
    layers.MaxPooling2D(),
    layers.Flatten(),
    layers.Dense(128, activation='relu'),
    layers.Dense(numClasses)
    ])
    

 

Answer 1

如果x.shape 是(1,224,224,3)，我检查了您的代码并没有发现问题。

>>> from tensorflow.keras import layers, Sequential    
>>> model=Sequential([
        layers.experimental.preprocessing.Rescaling(1./255,input_shape=(imgSize,imgSize,3)),
        layers.Conv2D(16,3,padding='same',activation='relu'),
        layers.MaxPooling2D(),
        layers.Conv2D(32,3,padding='same',activation='relu'),
        layers.MaxPooling2D(),
        layers.Conv2D(64,3,padding='same',activation='relu'),
        layers.MaxPooling2D(),
        layers.Flatten(),
        layers.Dense(128, activation='relu'),
        layers.Dense(numClasses)
        ])

    >>> import numpy as np
    >>> x = np.zeros((1,224,224,3))
    >>> model.predict(x)
    2021-05-02 18:35:15.325046: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)
    array([[0., 0.]], dtype=float32)
    >>>

请检查您的 x.shape，如果它与 input_shape 不兼容，您将收到此错误！

这是模型摘要：

>>> model.summary()
Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
rescaling_2 (Rescaling)      (None, 224, 224, 3)       0
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 224, 224, 16)      448
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 112, 112, 16)      0
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 112, 112, 32)      4640
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 56, 56, 32)        0
_________________________________________________________________
conv2d_8 (Conv2D)            (None, 56, 56, 64)        18496
_________________________________________________________________
max_pooling2d_8 (MaxPooling2 (None, 28, 28, 64)        0
_________________________________________________________________
flatten_2 (Flatten)          (None, 50176)             0
_________________________________________________________________
dense_3 (Dense)              (None, 128)               6422656
_________________________________________________________________
dense_4 (Dense)              (None, 2)                 258
=================================================================
Total params: 6,446,498
Trainable params: 6,446,498
Non-trainable params: 0
_________________________________________________________________
>>>