I have an audio dataset. I have process it and built an array as:
(audio Files, number of features, feature vector, 1)=(18, 64, 1688, 1)
But when I pass this to my model (file attach below), It shows me following error:
ValueError: Input 0 of layer sequential_54 is incompatible with the layer: expected ndim=5, found ndim=4. Full shape received: (None, 64, 1688, 1)
And when I again reshape it to get ndim=5 as:
(audio Files, number of features, feature vector, 1, 1)=(18, 64, 1688, 1, 1)
It again gives me an error says that:
Input 0 of layer conv_lst_m2d_83 is incompatible with the layer: expected ndim=5, found ndim=6. Full shape received: (None, None, 64, 211, 1, 128)
I do not know how to resolve this issue, and I have follow this link that https://stackoverflow.com/a/47665700/12559855 but my issue is not resolved.
Kindly guide me in this regard. It will be a great favor.
Honestly, I'm not familiar with how to structure audio input for machine learning, so take my answer with many grains of salt.
However, having dimensions of shape 1 in your input shape is essentially useless and meaningless; they're not true dimensions.
How many "spatial" dimensions does your audio input have?
From the Keras documentation:
Consider a batch of 32 video samples, where each sample is a 128x128 RGB image with channels_last data format, across 10 timesteps. The batch input shape is (32, 10, 128, 128, 3)
Your model code would then be:
inputs = tf.keras.Input(shape=(10, 128, 128, 3)) conv_2d_layer = tf.keras.layers.Conv2D(64, (3, 3)) outputs = tf.keras.layers.TimeDistributed(conv_2d_layer)(inputs)
This is the Functional model construction rather than Sequential, but I recommend learning to use it. At its simplest, it's only a slight change from sequential, but affords a lot more control.
- Badges
- Science topic
- Similar topics
- Filing