0

I had a similar problem to the one given in the following link:

ValueError: `decode_predictions` expects a batch of predictions (i.e. a 2D array of shape (samples, 1000)). Found array with shape: (1, 7)

I have been able to successfully run my code and also got the probabilities for each class.My concern is what will be the order of the output probabilities?Meaning how will we know which probability belongs to which class since we are assigning or mapping them manually.any assistance would be really appreciated!

Code:

class_list=style_info['articleType'].unique()
base_model = ResNet50(weights='imagenet',include_top=False,input_shape=(80, 60, 3))
img=glob.glob('images\\'+str(style_info.iloc[55,0])+'.jpg')
# add a global spatial average pooling layer
x = base_model.output
x = GlobalAveragePooling2D()(x)
# add a fully-connected layer
x = Dense(1024, activation='relu')(x)
# and a logistic layer -- let's say we have 7 classes
predictions = Dense(no_of_classes, activation='softmax')(x) 
res_model = Model(inputs=base_model.input, outputs=predictions)
image1 = image.load_img(img[0], target_size=(80,60))
x = image.img_to_array(image1)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
preds = res_model.predict(x)
# decode the results into a list of tuples (class, description, probability)
# (one such list for each sample in the batch)
#print('Predicted:', decode_predictions(preds, top=5)[0])
# Predicted: [(u'n02504013', u'Indian_elephant', 0.82658225), (u'n01871265', u'tusker', 0.1122357), (u'n02504458', u'African_elephant', 0.061040461)]
class_list[np.argmax(preds[0])]
Shrinidhi Narasimhan
  • 113
  • 1
  • 10
  • The order of classes in the output is whatever the order of classes is in your training data – FlyingTeller Sep 03 '19 at 12:56
  • But I don't have a specific order.I have a column that gives the articleType of each row item.The unique values of this column will be all the classes.So i have used nunique() function on the column and obtained the list of classes.There is no order in this now. – Shrinidhi Narasimhan Sep 03 '19 at 13:05
  • And I have not supplied any training data to resnet.I have directly used one image and tried to predict.Please note the updated question for the code. – Shrinidhi Narasimhan Sep 03 '19 at 13:07

0 Answers0