Keras: Model Predict, Error When Checking Input Shape

Question

I have a Keras model and a numpy array that I want to call predict on. Specifically, I have:

a numpy.ndarray called test that looks like this:

array([    0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0,     0,     0,     0,     0,     0,     0,
       0,     0,     0, 12920, 11891,  4605,  2425,  6780,  5096,
   13821,  4405, 10345,  4468,  5910, 11891, 10906, 14994, 12073,
    8581,  3544, 13846,  3110,  2425,  3407,  9631, 13846,  4479,
    9964,  2556,  4479,  2686,  8895, 10959,  1531, 11891,  1494,
   10376, 13846, 12856, 13846,  3110,  2425,  3407,  3267,   181,
    4479, 14842,  4639,  7723, 11891, 11449,  2425,  5662,  2282,
    5129,  2518, 13846,  4479,  4780,  2598,  4926,   543,  7304,
   12020,  8143, 10998, 13846, 12853, 13846, 12856, 11891,  3785,
    9131,  7448, 13846, 10376, 13846,  8245,  3788, 12211,  2425,
   13614, 10049,  2556,  8245,  1406,  6423,  3110,  2425,  3407,
    5726,  2619,  1494, 13694,  7434, 12086,  7304,  3267,  9184])

When I do test.shape, I see:

(180,)

When I do model.predict(test), I get back:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "...python2.7/site-packages/keras/engine/training.py", line 1152, in predict
    x, _, _ = self._standardize_user_data(x)
  File "...python2.7/site-packages/keras/engine/training.py", line 754, in _standardize_user_data
    exception_prefix='input')
  File "...python2.7/site-packages/keras/engine/training_utils.py", line 136, in standardize_input_data
    str(data_shape))
ValueError: Error when checking input: expected sequence to have shape (180,) but got array with shape (1,)

Any idea what I'm doing incorrectly?

Thanks!

Answer 1

Keras wants batched data, so here are some tips about the shapes of data you should know:

The shapes of data should be:

• for vector data: (samples, features)
• for time series data: (samples, timesteps, features)
• for image data: (samples, height, width, channels) or (samples, channels, height, width)
• for video data: (samples, frames, height, width, channels) or (samples, frames, channels, height, width)

So your test data should have the shape of (1, 180). Just reshape your data:

test.reshape(1,-1)

Another thing that you should know is that when Keras wants a specific shape, it means the shape of one sample, not including the first dimension of your real input data: (samples,). But your data should have the samples dimension. And Keras will handle the first dimension which is the number of samples by itself.