definition.yaml

benchmark:
  Google Speech Commands test set:
    Accuracy: 0.958
description: This is a fully quantized version (asymmetrical int8) of the MicroNet
  Medium model developed by Arm, from the MicroNets paper. This model is trained on
  the 'Google Speech Commands' dataset.
license:
- Apache-2.0
network:
  file_size_bytes: 181968
  filename: kws_micronet_m.tflite
  framework: TensorFlow Lite
  hash:
    algorithm: sha1
    value: fd03a6b24548ea99cf487dbd682937df5718cef1
  provenance: https://arxiv.org/pdf/2010.11267.pdf
network_parameters:
  input_nodes:
  - description: A one second audio clip, converted to a 2D MFCC computed from a speech
      frame of length 40ms and stride 20ms.
    example_input:
      path: models/keyword_spotting/micronet_medium/tflite_int8/testing_input/input
    name: input
    shape:
    - 1
    - 49
    - 10
    - 1
  output_nodes:
  - description: The probability on 12 keywords.
    name: Identity
    shape:
    - 1
    - 12
    test_output_path: models/keyword_spotting/micronet_medium/tflite_int8/testing_output/Identity
operators:
  TensorFlow Lite:
  - AVERAGE_POOL_2D
  - CONV_2D
  - DEPTHWISE_CONV_2D
  - RELU6
  - RESHAPE
paper: https://arxiv.org/pdf/2010.11267.pdf