CNNの保存済みモデルをロードしてpredictしたい

前提・実現したいこと

CNNの保存済みモデルをロードして、predictをしたいです。

発生している問題・エラーメッセージ

ValueError: Input 0 is incompatible with layer model: expected shape=(None, 120, 90, 1), found shape=(None, 90, 1)

該当のソースコード

python
1#X_testはnumpy形式で形状は,(120, 90, 1)です。（確認済み）
2loaded_model = load_model('model.h5', custom_objects=ak.CUSTOM_OBJECTS)
3predicted_y = loaded_model.predict(X_test)

モデルの構造

Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 120, 90, 1)]      0         
_________________________________________________________________
cast_to_float32 (CastToFloat (None, 120, 90, 1)        0         
_________________________________________________________________
normalization (Normalization (None, 120, 90, 1)        3         
_________________________________________________________________
conv2d (Conv2D)              (None, 118, 88, 32)       320       
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 116, 86, 64)       18496     
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 58, 43, 64)        0         
_________________________________________________________________
dropout (Dropout)            (None, 58, 43, 64)        0         
_________________________________________________________________
flatten (Flatten)            (None, 159616)            0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 159616)            0         
_________________________________________________________________
dense (Dense)                (None, 3)                 478851    
_________________________________________________________________
classification_head_1 (Softm (None, 3)                 0