tflite_int8

MicroNet Small INT8

Description

This is a fully quantized version (asymmetrical int8) of the MicroNet Small model developed by Arm, from the MicroNets paper. This model is trained on the 'Google Speech Commands' dataset.

License

Apache-2.0

Related Materials

Class Labels

The class labels associated with this model can be downloaded by running the script get_class_labels.sh.

Network Information

Network Information	Value
Framework	TensorFlow Lite
SHA-1 Hash	d13a25dbe34a0f2758879b9d3e7c7bad94c68ec7
Size (Bytes)	114512
Provenance	https://arxiv.org/pdf/2010.11267.pdf
Paper	https://arxiv.org/pdf/2010.11267.pdf

Performance

Platform	Optimized
Cortex-A	✖️
Cortex-M	✔️
Mali GPU	✔️
Ethos U	✔️

Key

• ✔️ - Will run on this platform.
• ✖️ - Will not run on this platform.

Accuracy

Dataset: Google Speech Commands Test Set

Metric	Value
Accuracy	0.953

Optimizations

Optimization	Value
Quantization	INT8

Network Inputs

Input Node Name	Shape	Description
input	(1, 49, 10, 1)	A one second audio clip, converted to a 2D MFCC computed from a speech frame of length 40ms and stride 20ms.

Network Outputs

Output Node Name	Shape	Description
Identity	(1, 12)	The probability on 12 keywords.