Ex Machina / Keyword_Spotting_Bionic_Hand Public
Target: Arduino Nano 33 BLE Sense (Cortex-M4F 64MHz) Clone this project
Experiments EON Tuner
The EON Tuner helps you quickly run hyper-parameter sweeps that explore different pre-processing + model architectures optimized for your defined objectives. Clone this project to use the EON Tuner.

Target

No name set

Arduino Nano 33 BLE Sense (Cortex-M4F 64MHz)

100 ms

256 kB

1024 kB

Filters

Status

DSP type

Model type

View

Data set

Variant

Sort

General

F1-score

Precision

Recall

98%
mfe-conv1d-99f
PERFORMANCE
LATENCY
194 ms of 100 ms
Exceeds target by 94 ms
RAM
18 kB of 256 kB
ROM
41 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.02 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 98%

Type Filters Kernel Rate
conv1d 32 3 -
conv1d 64 3 -
dropout - - 0.25

3/28/2023, 6:17:23 PM

98%
mfe-conv1d-a9d
PERFORMANCE
LATENCY
86 ms of 100 ms
RAM
14 kB of 256 kB
ROM
30 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.05 | 0.05 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 98%

Type Filters Kernel Rate
conv1d 16 3 -
conv1d 32 3 -
dropout - - 0.25

3/28/2023, 6:26:29 PM

98%
mfe-conv1d-269
PERFORMANCE
LATENCY
221 ms of 100 ms
Exceeds target by 121 ms
RAM
22 kB of 256 kB
ROM
67 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.05 | 0.025 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 98%

Type Filters Kernel Rate
conv1d 32 3 -
conv1d 64 3 -
conv1d 128 3 -
dropout - - 0.5

3/28/2023, 6:30:25 PM

98%
mfe-conv1d-546
PERFORMANCE
LATENCY
214 ms of 100 ms
Exceeds target by 114 ms
RAM
22 kB of 256 kB
ROM
67 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.02 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 98%

Type Filters Kernel Rate
conv1d 16 3 -
conv1d 32 3 -
conv1d 64 3 -
conv1d 128 3 -
dropout - - 0.25

3/28/2023, 6:38:17 PM

97%
mfe-conv1d-93c
PERFORMANCE
LATENCY
191 ms of 100 ms
Exceeds target by 91 ms
RAM
20 kB of 256 kB
ROM
40 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.02 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 97%

Type Filters Kernel Rate
Data augmentation
conv1d 16 3 -
conv1d 32 3 -
conv1d 64 3 -
dropout - - 0.25

3/28/2023, 6:34:00 PM

97%
mfe-conv1d-83f
PERFORMANCE
LATENCY
353 ms of 100 ms
Exceeds target by 253 ms
RAM
27 kB of 256 kB
ROM
34 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.01 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 97%

Type Filters Kernel Rate
conv1d 16 3 -
conv1d 32 3 -
dropout - - 0.25

3/28/2023, 6:45:14 PM

95%
mfe-conv1d-db8
PERFORMANCE
LATENCY
202 ms of 100 ms
Exceeds target by 102 ms
RAM
18 kB of 256 kB
ROM
41 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.02 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 95%

Type Filters Kernel Rate
conv1d 32 3 -
conv1d 64 3 -
dropout - - 0.5

3/28/2023, 6:03:05 PM

94%
mfe-conv1d-003
PERFORMANCE
LATENCY
341 ms of 100 ms
Exceeds target by 241 ms
RAM
28 kB of 256 kB
ROM
32 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.02 | 0.01 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 94%

Type Filters Kernel Rate
conv1d 8 3 -
conv1d 16 3 -
conv1d 32 3 -
dropout - - 0.25

3/28/2023, 6:33:19 PM

94%
mfe-conv1d-f32
PERFORMANCE
LATENCY
164 ms of 100 ms
Exceeds target by 64 ms
RAM
23 kB of 256 kB
ROM
168 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.05 | 0.05 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 94%

Type Filters Kernel Rate
conv1d 32 3 -
conv1d 64 3 -
conv1d 128 3 -
conv1d 256 3 -
dropout - - 0.25

3/28/2023, 6:42:51 PM

92%
mfe-mobilenetv1-839
PERFORMANCE
LATENCY
423 ms of 100 ms
Exceeds target by 323 ms
RAM
65 kB of 256 kB
ROM
108 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
500 ms |
Enabled

MFE

0.02 | 0.01 | 40

ACCURACY (KERAS-TRANSFER-KWS)
TRANSFER LEARNING (KEYWORD SPOTTING)

0.01 | 30 | 92%

MobileNetV1 0.1
0 | 0.1

3/28/2023, 6:07:17 PM

90%
mfcc-conv1d-691
PERFORMANCE
LATENCY
618 ms of 100 ms
Exceeds target by 518 ms
RAM
31 kB of 256 kB
ROM
41 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFCC

0.02 | 0.01 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 90%

Type Filters Kernel Rate
Data augmentation
conv1d 16 3 -
conv1d 32 3 -
conv1d 64 3 -
dropout - - 0.5

3/28/2023, 6:28:09 PM

90%
mfe-conv1d-989
PERFORMANCE
LATENCY
161 ms of 100 ms
Exceeds target by 61 ms
RAM
19 kB of 256 kB
ROM
30 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFE

0.05 | 0.025 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 90%

Type Filters Kernel Rate
Data augmentation
conv1d 8 3 -
conv1d 16 3 -
conv1d 32 3 -
dropout - - 0.5

3/28/2023, 6:41:58 PM

90%
mfcc-conv1d-642
PERFORMANCE
LATENCY
166 ms of 100 ms
Exceeds target by 66 ms
RAM
15 kB of 256 kB
ROM
29 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFCC

0.05 | 0.05 | 40

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 90%

Type Filters Kernel Rate
conv1d 16 3 -
conv1d 32 3 -
dropout - - 0.25

3/28/2023, 6:18:49 PM

89%
mfcc-conv1d-de8
PERFORMANCE
LATENCY
500 ms of 100 ms
Exceeds target by 400 ms
RAM
18 kB of 256 kB
ROM
0 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFCC

0.032 | 0.016 | 40

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 89%

Type Filters Kernel Rate
conv1d 16 3 -
conv1d 32 3 -
conv1d 64 3 -
dropout - - 0.25

3/28/2023, 6:19:36 PM

88%
mfcc-conv2d-0d9
PERFORMANCE
LATENCY
171 ms of 100 ms
Exceeds target by 71 ms
RAM
19 kB of 256 kB
ROM
52 kB of 1024 kB
DSP NN Unused
TIME-SERIES INPUT

1000 ms |
1000 ms |
Enabled

MFCC

0.05 | 0.05 | 32

ACCURACY (KERAS)
CLASSIFICATION

0.005 | 100 | 88%

Type Filters Kernel Rate
conv2d 8 3 -
conv2d 16 3 -
conv2d 32 3 -
conv2d 64 3 -
dropout - - 0.5

3/28/2023, 6:25:24 PM

Help & documentation Get help on the forum Send feedback Contact sales
Join us
Terms of Service Privacy policy