模型预测输入形状不兼容求助：预期(None,150,80,1)获(None,80,1)

模型预期输入形状为(batchSize, imageWidth, imageHeight, imgChannels)即(1,150,80,1)，但实际检测到的形状为(None,80,1)，无法正常调用prediction_model.predict()。

错误信息 #

ValueError: Input 0 of layer "model" is incompatible with the layer: expected shape=(None, 150, 80, 1), found shape=(None, 80, 1)

预处理代码 #

# Create a tensor from an image

# 1. Read the image
img = tf.io.read_file('/content/drive/MyDrive/testImg.jpg')
# 2. Decode and convert to grayscale
img = tf.io.decode_jpeg(img, channels=1)
#3. Convert to float32 in the range [0, 1]
img = tf.image.convert_image_dtype(img, tf.float32)
# 4. Change the size to the desired size
img = tf.image.resize(img, [80, 150])

print(img.shape)

#5. Move
img = tf.transpose(img, perm=[1, 0, 2])

print(img.shape)

pred = prediction_model.predict(img)
pred_texts = decode_batch_predictions(pred)

控制台输出 #

(80, 150, 1)
(150, 80, 1)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-19-e1b6d597bdc5> in <module>
     14 print(img.shape)
     15 
---> 16 preds = prediction_model.predict(img)
     17 #pred_texts = decode_batch_predictions(preds)

1 frames
/usr/local/lib/python3.8/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs)
     65     except Exception as e:  # pylint: disable=broad-except
     66       filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67       raise e.with_traceback(filtered_tb) from None
     68     finally:
     69       del filtered_tb

/usr/local/lib/python3.8/dist-packages/keras/engine/training.py in tf__predict_function(iterator)
     13                 try:
     14                     do_return = True
---> 15                     retval_ = ag__.converted_call(ag__.ld(step_function), (ag__.ld(self), ag__.ld(iterator)), None, fscope)
     16                 except:
     17                     do_return = False

ValueError: in user code:

    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1845, in predict_function  *
        return step_function(self, iterator)
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1834, in step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1823, in run_step  **
        outputs = model.predict_step(data)
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/training.py", line 1791, in predict_step
        return self(x, training=False)
    File "/usr/local/lib/python3.8/dist-packages/keras/utils/traceback_utils.py", line 67, in error_handler
        raise e.with_traceback(filtered_tb) from None
    File "/usr/local/lib/python3.8/dist-packages/keras/engine/input_spec.py", line 264, in assert_input_compatibility
        raise ValueError(f'Input {input_index} of layer "{layer_name}" is '

    ValueError: Input 0 of layer "model" is incompatible with the layer: expected shape=(None, 150, 80, 1), found shape=(None, 80, 1)

模型摘要（prediction_model.summary()） #

Model: "model_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 image (InputLayer)          [(None, 150, 80, 1)]      0         
                                                                 
 Conv1 (Conv2D)              (None, 150, 80, 32)       320       
                                                                 
 pool1 (MaxPooling2D)        (None, 75, 40, 32)        0         
                                                                 
 Conv2 (Conv2D)              (None, 75, 40, 64)        18496     
                                                                 
 pool2 (MaxPooling2D)        (None, 37, 20, 64)        0         
                                                                 
 reshape (Reshape)           (None, 37, 1280)          0         
                                                                 
 dense1 (Dense)              (None, 37, 64)            81984     
                                                                 
 dropout (Dropout)           (None, 37, 64)            0         
                                                                 
 bidirectional (Bidirectiona  (None, 37, 256)          197632    
 l)                                                               
                                                                 
 bidirectional_1 (Bidirectio  (None, 37, 128)          164352    
 nal)                                                             
                                                                 
 dense2 (Dense)              (None, 37, 12)            1548      
                                                                 
=================================================================
Total params: 464,332
Trainable params: 464,332
Non-trainable params: 0
_________________________________________________________________

疑问 #

请问我哪里处理有误？为何无法将预处理后的单张图像传入prediction_model.predict()？

问题出在缺少batch维度：

预处理后得到的img形状是(150,80,1)，但Keras的predict()方法要求输入必须包含batch维度（即第一维是样本数）。模型预期的输入形状(None,150,80,1)中，None代表可变的batch大小，单张图片对应的batch大小是1，所以需要把输入形状变成(1,150,80,1)。

修改方法：在调用predict()之前，给img增加一个batch维度：

# 在transpose之后添加这行代码
img = tf.expand_dims(img, axis=0)
print(img.shape)  # 此时输出应为(1,150,80,1)

pred = prediction_model.predict(img)
pred_texts = decode_batch_predictions(pred)

另外验证一下resize和transpose步骤：

• 先用tf.image.resize(img, [80,150])把图像改成(80,150,1)，再用transpose(perm=[1,0,2])转成(150,80,1)，这一步是正确的，符合模型要求的宽高顺序。

总结：Keras模型的predict()、fit()等方法都要求输入是批量数据，哪怕只有一张图片，也要手动添加batch维度。

内容的提问来源于stack exchange，提问作者Vipz