ValueError：layersequence_4 的输入 0 与 layer 不兼容：：预期 min_ndim=4，发现 ndim=3。 收到的完整形状：（32、224、3）

Question

I am getting this error on using model.predict() on local images. It is trained on shape (224,224,3) and I've resized the input image to (224,224,3) still it shows error. The images with same sshape in test array are getting predicted without any issues. I am new to this, can someone please tell the mistake.

#imports

data_dir = "/content/gdrive/MyDrive/mask_detector"
import pathlib
data_dir = pathlib.Path(data_dir)

mask_data_dict = {'mask' : list(data_dir.glob('Mask/*')),'non_mask' : list(data_dir.glob('Non_Mask/*'))}
mask_labels_dict = {'mask' : 0,'non_mask' : 1}

X,y = [], []

for mask_name,images in mask_data_dict.items():
  for image in images:
    img_array = cv2.imread(str(image))
    img_array_resized = cv2.resize(img_array,(224,224))
    X.append(img_array_resized)
    y.append(mask_labels_dict[mask_name])

X = np.array(X)
y = np.array(y)

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X,y,random_state= 0)

X_test_scaled = X_test/255
X_train_scaled = X_train/255

X_test = X_test.astype('float32')
X_train = X_train.astype('float32')

y_train_categorical = keras.utils.to_categorical(y_train, num_classes = 2, dtype = 'float32')
y_test_categorical = keras.utils.to_categorical(y_test, num_classes = 2, dtype = 'float32')

model = keras.Sequential([
        keras.layers.Conv2D(16,(3,3), padding='same',activation = 'relu',input_shape=(224,224,3)),
        keras.layers.MaxPooling2D(),
        keras.layers.Conv2D(32,(3,3), padding='same',activation = 'relu'),
        keras.layers.MaxPooling2D(),
        keras.layers.Conv2D(64,(3,3), padding='same',activation = 'relu'),
        keras.layers.Flatten(),
        keras.layers.Dense(128, activation = 'relu'),
        keras.layers.Dense(2, activation = 'sigmoid')])
model.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'])
model.fit(X_train_scaled, y_train_categorical, epochs = 5)

def prediction(image):
    pred_array = cv2.imread(image)
    pred_array = np.array(pred_array)
    pred_array_resized = cv2.resize(pred_array,(224,224,))
    pred_array_scaled = np.array(pred_array_resized)/255
    model.predict(pred_array_scaled)
    
    
prediction("example.jpeg") #shape 
#ValueError: Input 0 of layer sequential_4 is incompatible with the layer: : expected min_ndim=4, found ndim=3. Full shape received: (32, 224, 3)

classes = ["mask","non-mask"]

cv2_imshow(X_test[9])
classes[np.argmax(model.predict(X_test_scaled)[9])] #this works fine

Answer 1

Model needs 4-dim array, so before predicting, add this line to expand dim to NHWC like.

pred_array_scaled = np.expand_dims(pred_array_scaled, axis=0)