Skip to content
This repository was archived by the owner on Jul 18, 2025. It is now read-only.

Commit b9e26e6

Browse files
orly-armorly-arm
authored
Merge pull request #44 from ARM-software/kws_models_update
KWS models - update
2 parents 7d96402 + 48a22ee commit b9e26e6

38 files changed

Lines changed: 100 additions & 92 deletions

File tree

‎README.md‎

Lines changed: 6 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,3 @@
1-
21
# Model Zoo
32
![version](https://img.shields.io/badge/version-21.08-0091BD)
43
> A collection of machine learning models optimized for Arm IP.
@@ -109,7 +108,7 @@
109108
<td align="center">:heavy_check_mark: </td>
110109
<td align="center">:heavy_check_mark: </td>
111110
<td align="center">:heavy_check_mark: </td>
112-
<td align="center">0.929</td>
111+
<td align="center">0.931</td>
113112
</tr>
114113
<tr>
115114
<td><a href="models/keyword_spotting/cnn_medium/tflite_int8">CNN Medium INT8 *</a></td>
@@ -119,7 +118,7 @@
119118
<td align="center">:heavy_check_mark: </td>
120119
<td align="center">:heavy_check_mark: </td>
121120
<td align="center">:heavy_check_mark: </td>
122-
<td align="center">0.913</td>
121+
<td align="center">0.911</td>
123122
</tr>
124123
<tr>
125124
<td><a href="models/keyword_spotting/cnn_small/tflite_int8">CNN Small INT8 *</a></td>
@@ -129,7 +128,7 @@
129128
<td align="center">:heavy_check_mark: </td>
130129
<td align="center">:heavy_check_mark: </td>
131130
<td align="center">:heavy_check_mark: </td>
132-
<td align="center">0.914</td>
131+
<td align="center">0.912</td>
133132
</tr>
134133
<tr>
135134
<td><a href="models/keyword_spotting/dnn_large/tflite_int8">DNN Large INT8 *</a></td>
@@ -149,7 +148,7 @@
149148
<td align="center">:heavy_check_mark: </td>
150149
<td align="center">:heavy_check_mark: </td>
151150
<td align="center">:heavy_check_mark: </td>
152-
<td align="center">0.846</td>
151+
<td align="center">0.844</td>
153152
</tr>
154153
<tr>
155154
<td><a href="models/keyword_spotting/dnn_small/tflite_int8">DNN Small INT8 *</a></td>
@@ -159,7 +158,7 @@
159158
<td align="center">:heavy_check_mark: </td>
160159
<td align="center">:heavy_check_mark: </td>
161160
<td align="center">:heavy_check_mark: </td>
162-
<td align="center">0.827</td>
161+
<td align="center">0.825</td>
163162
</tr>
164163
<tr>
165164
<td><a href="models/keyword_spotting/ds_cnn_large/tflite_clustered_fp32">DS-CNN Clustered FP32 *</a></td>
@@ -199,7 +198,7 @@
199198
<td align="center">:heavy_check_mark: HERO</td>
200199
<td align="center">:heavy_check_mark: </td>
201200
<td align="center">:heavy_check_mark: </td>
202-
<td align="center">0.934</td>
201+
<td align="center">0.941</td>
203202
</tr>
204203
<tr>
205204
<td><a href="models/keyword_spotting/ds_cnn_small/tflite_int8">DS-CNN Small INT8 *</a></td>

‎models/keyword_spotting/cnn_large/tflite_int8/README.md‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -17,8 +17,8 @@ Code to recreate this model can be found here: https://github.com/ARM-software/M
1717
| Network Information | Value |
1818
|---------------------|------------------|
1919
| Framework | TensorFlow Lite |
20-
| SHA-1 Hash | 59e6986c3eca496fa3d54176ac66bb7dc9ff36e0 |
21-
| Size (Bytes) | 486560 |
20+
| SHA-1 Hash | a61ab748ae8f52f78ab568342db67a792c6ecf34 |
21+
| Size (Bytes) | 484600 |
2222
| Provenance | https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m |
2323
| Paper | https://arxiv.org/abs/1711.07128 |
2424

@@ -27,7 +27,7 @@ Dataset: Google Speech Commands Test Set
2727

2828
| Metric | Value |
2929
|--------|-------|
30-
| Accuracy | 0.929 |
30+
| Accuracy | 0.931 |
3131

3232
## Performance
3333
| Platform | Optimized |

‎models/keyword_spotting/cnn_large/tflite_int8/cnn_l_quantized.tflite‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:db7e359e766f96bcafee00ff475672cd0268cdb84156bfdc2b05d70ab24dc32a
3-
size 486560
2+
oid sha256:7dee2cb5f152bda13e09593b2355c174c012080b569ba26bce1f4132feb44633
3+
size 484600

‎models/keyword_spotting/cnn_large/tflite_int8/definition.yaml‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,20 @@
11
benchmark:
22
Google Speech Commands test set:
3-
Accuracy: 0.929
3+
Accuracy: 93.09%
44
description: 'This is a fully quantized version (asymmetrical int8) of the CNN Large
55
model developed by Arm, with training checkpoints, from the Hello Edge paper. Code
66
to recreate this model can be found here: https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m'
77
license:
88
- Apache-2.0
99
network:
10-
file_size_bytes: 486560
10+
file_size_bytes: 484600
1111
filename: cnn_l_quantized.tflite
1212
framework: TensorFlow Lite
1313
hash:
1414
algorithm: sha1
15-
value: 59e6986c3eca496fa3d54176ac66bb7dc9ff36e0
15+
value: a61ab748ae8f52f78ab568342db67a792c6ecf34
1616
provenance: https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m
17+
quality_level: null
1718
network_parameters:
1819
input_nodes:
1920
- description: The input is a processed MFCCs of shape (1, 490)

‎models/keyword_spotting/cnn_large/tflite_int8/testing_input/input/0.npy‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:c617e49b3881660f02eb118908b915686b4c9018e0904cc4015731da5b8f083b
3-
size 2088
2+
oid sha256:7f12421d1892c3c971f2cde7e9effb1334d0ebf4dc408dc979d3418f49d66bda
3+
size 618

‎models/keyword_spotting/cnn_large/tflite_int8/testing_output/Identity/0.npy‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:263b365ec4a0059ec477dc4bcbfec16a49b7c2920641ed48efa201c848424cc4
3-
size 176
2+
oid sha256:bb6ae4cf97cee9fea31a725d8f61ae0ca3d1b7b5059bcede551f365f94f8feaf
3+
size 140

‎models/keyword_spotting/cnn_medium/tflite_int8/README.md‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -17,8 +17,8 @@ Code to recreate this model can be found here: https://github.com/ARM-software/M
1717
| Network Information | Value |
1818
|---------------------|------------------|
1919
| Framework | TensorFlow Lite |
20-
| SHA-1 Hash | 389c6c2c7d289c0018e2dabcc66271811e52874c |
21-
| Size (Bytes) | 187840 |
20+
| SHA-1 Hash | 6bc68074d960bbb0c695e19fd96fd7903131ef60 |
21+
| Size (Bytes) | 186064 |
2222
| Provenance | https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m |
2323
| Paper | https://arxiv.org/abs/1711.07128 |
2424

@@ -27,7 +27,7 @@ Dataset: Google Speech Commands Test Set
2727

2828
| Metric | Value |
2929
|--------|-------|
30-
| Accuracy | 0.913 |
30+
| Accuracy | 0.911 |
3131

3232
## Performance
3333
| Platform | Optimized |

‎models/keyword_spotting/cnn_medium/tflite_int8/cnn_m_quantized.tflite‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:c5b248c9bbd1d49e2887eb410853338838debd8465363aebf77ed66601f6d279
3-
size 187840
2+
oid sha256:ae3ddb36a6a397b4394122abf320f92754f148a432e9a6de12619e78d3d853c1
3+
size 186064

‎models/keyword_spotting/cnn_medium/tflite_int8/definition.yaml‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,20 @@
11
benchmark:
22
Google Speech Commands test set:
3-
Accuracy: 0.913
3+
Accuracy: 91.08%
44
description: 'This is a fully quantized version (asymmetrical int8) of the CNN Medium
55
model developed by Arm, with training checkpoints, from the Hello Edge paper. Code
66
to recreate this model can be found here: https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m'
77
license:
88
- Apache-2.0
99
network:
10-
file_size_bytes: 187840
10+
file_size_bytes: 186064
1111
filename: cnn_m_quantized.tflite
1212
framework: TensorFlow Lite
1313
hash:
1414
algorithm: sha1
15-
value: 389c6c2c7d289c0018e2dabcc66271811e52874c
15+
value: 6bc68074d960bbb0c695e19fd96fd7903131ef60
1616
provenance: https://github.com/ARM-software/ML-examples/tree/master/tflu-kws-cortex-m
17+
quality_level: null
1718
network_parameters:
1819
input_nodes:
1920
- description: The input is a processed MFCCs of shape (1, 490)

‎models/keyword_spotting/cnn_medium/tflite_int8/testing_input/input/0.npy‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:e9e78e930e911e01c9de252a2ce4bd0a24d24b92df99360d22a8ce4aa0fa9374
3-
size 2088
2+
oid sha256:d2407c02e6ffdd0b88ee777cc692034d5f9202d1c130ce3e87405664bdd265ed
3+
size 618

0 commit comments

Comments
 (0)