How to solve, tensorflow.python.framework.errors_impl.InvalidArgumentError?

Question

import tensorflow as tf
import numpy as np
from sklearn.model_selection import train_test_split

np.random.seed(4213)

data = np.random.randint(low=1,high=29, size=(500, 160, 160, 10)) 
labels = np.random.randint(low=0,high=5, size=(500, 160, 160)) 
nclass = len(np.unique(labels))
print (nclass)

samples, width, height, nbands = data.shape


X_train, X_test, y_train, y_test = train_test_split(data, labels, test_size=0.25, random_state=421)

print (X_train.shape)
print (y_train.shape)

arch = tf.keras.applications.VGG16(input_shape=[width, height, nbands],
                      include_top=False,
                      weights=None)

model = tf.keras.Sequential()
model.add(arch)
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(nclass))

model.compile(optimizer = tf.keras.optimizers.Adam(0.0001),
              loss=tf.keras.losses.SparseCategoricalCrossentropy(),
              metrics=[tf.keras.metrics.SparseCategoricalAccuracy()])
    

model.fit(X_train,
          y_train,                                 
          epochs=3,
          batch_size=32,
          verbose=2)


res = model.predict(X_test)
print(res.shape)

When running the above code for semantic segmentation I get Exception has occurred:

InvalidArgumentError
 Incompatible shapes: [32,160,160] vs. [32]
     [[node Equal (defined at c...:38) ]] [Op:__inference_train_function_1815]


tensorflow.python.framework.errors_impl.InvalidArgumentError

You should post the complete error message when dealing with errors in Stackoverflow — ibarrond
– ibarrond, Commented Jul 28, 2020 at 14:35
As mentioned in the posted answer, your data is for image segmentation, but your model has the structure for image classification. You should pick a model for that purpose, not VGG16. There are many papers out there discussing neural image segmentation (e.g. see here). — javidcf
– javidcf, Commented Jul 31, 2020 at 12:03

ibarrond · Accepted Answer · 2020-08-04 11:53:01Z

1

+50

Your issue comes from the size of the last layer (to avoid these mistakes it is always desirable to use python constants for N_IMAGES, WIDTH, HEIGHT, N_CHANNELS and N_CLASSES):

For image classification

You should assign one single label to each image. Try switching labels:

import tensorflow as tf
import numpy as np
from sklearn.model_selection import train_test_split

np.random.seed(4213)

N_IMAGES, WIDTH, HEIGHT, N_CHANNELS = (500, 160, 160, 10)
N_CLASSES  = 5

data = np.random.randint(low=1,high=29, size=(N_IMAGES, WIDTH, HEIGHT, N_CHANNELS)) 
labels = np.random.randint(low=0,high=N_CLASSES, size=(N_IMAGES)) 
#...

For semantic segmentation

Make sure your classifier (last layers of the network) is sized accordingly. In this case you need 1 class per pixel:

#...
model = tf.keras.Sequential()
model.add(arch)
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(width * height))
model.add(tf.keras.layers.Reshape([width , height]))
#...

This is the simplest you can get. Instead, you can set up multiple deconvolution layers to act as classifier, or you can even flip the arch architecture over and use it to generate the classification results. Orthogohally, you can perform one_hot encoding on the labels and thus expand them by a factor of N_CLASSES, effectively multiplying the number of neurons in the last layer.

edited Aug 4, 2020 at 11:53

answered Jul 28, 2020 at 15:01

ibarrond

7,8716 gold badges32 silver badges52 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

ibarrond Over a year ago

That's an entirely different question! You should either close this question and ask a new one, or modify it accordingly. Do you have any references for semantic segmentation?

Collectives™ on Stack Overflow

How to solve, tensorflow.python.framework.errors_impl.InvalidArgumentError?

1 Answer 1

For image classification

For semantic segmentation

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

For image classification

For semantic segmentation

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related