keyword_spotting/ds_cnn_small/model_package_tf/model_archive/TFLite/tflite_int16
This is a fully quantized int16 version of the DS-CNN Small model developed by Arm, from the Hello Edge paper.
Apache-2.0
Dataset Information
Value
Name
Google Speech Commands test set
Metric
Value
Accuracy
93.39%
HW Support
Value
Cortex-A
✔️
Cortex-M
✔️ HERO
Mali GPU
✔️
Ethos U
✔️
✔️ - Will run on this platform.
✖️ - Will not run on this platform.
Network Quality
Value
Recreate
✔️
Quality level
Hero
Vanilla
✔️
Clustered
✖️
Pruned
✖️
Quantization - default
✖️
Quantization - full
✔️
Input Node Name
Shape
Type
Example Path
Example Type
Example Shape
Example Use Case
serving_default_input:0
(1, 490)
int16
models/keyword_spotting/ds_cnn_small/model_package_tf/model_archive/TFLite/tflite_int16/testing_input
int16
[1, 490]
The input is a processed MFCCs
Output Node Name
Shape
Type
Example Path
Example Type
Example Shape
Example Use Case
StatefulPartitionedCall:0
(1, 12)
int16
models/keyword_spotting/ds_cnn_small/model_package_tf/model_archive/TFLite/tflite_int16/testing_output
int16
[1, 12]
The probability on 12 keywords