Edge Impulse Inc. / Cars binary classifier - EON Tuner Search Space Public
Target: Arduino Portenta H7 (Cortex-M7 480MHz) 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 Portenta H7 (Cortex-M7 480MHz)

100 ms

1024 kB

2048 kB

Filters

Status

DSP type

Model type

View

Data set

Variant

Sort

General

F1-score

Precision

Recall

98%
grayscale-mobilenetv2-1f2
PERFORMANCE
LATENCY
74 ms of 100 ms
RAM
288 kB of 1024 kB
ROM
199 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS-TRANSFER-IMAGE)
TRANSFER LEARNING (IMAGES)

0.0005 | 20 | 98%

MobileNetV2 0.05
16 | 0.5

9/28/2022, 8:02:09 AM

98%
grayscale-mobilenetv2-5dc
PERFORMANCE
LATENCY
84 ms of 100 ms
RAM
288 kB of 1024 kB
ROM
220 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS-TRANSFER-IMAGE)
TRANSFER LEARNING (IMAGES)

0.0005 | 20 | 98%

MobileNetV2 0.05
32 | 0.1

9/28/2022, 8:06:05 AM

97%
grayscale-mobilenetv1-351
PERFORMANCE
LATENCY
130 ms of 100 ms
Exceeds target by 30 ms
RAM
111 kB of 1024 kB
ROM
262 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS-TRANSFER-IMAGE)
TRANSFER LEARNING (IMAGES)

0.0005 | 20 | 97%

MobileNetV1 0.2
32 | 0.5 |

9/28/2022, 8:09:04 AM

97%
grayscale-mobilenetv1-5ea
PERFORMANCE
LATENCY
27 ms of 100 ms
RAM
65 kB of 1024 kB
ROM
140 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS-TRANSFER-IMAGE)
TRANSFER LEARNING (IMAGES)

0.0005 | 20 | 97%

MobileNetV1 0.1
16 | 0.5 |

9/28/2022, 8:04:21 AM

97%
grayscale-mobilenetv1-506
PERFORMANCE
LATENCY
25 ms of 100 ms
RAM
65 kB of 1024 kB
ROM
140 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS-TRANSFER-IMAGE)
TRANSFER LEARNING (IMAGES)

0.0005 | 20 | 97%

MobileNetV1 0.1
16 | 0.1 |

9/28/2022, 8:11:34 AM

84%
grayscale-conv2d-f56
PERFORMANCE
LATENCY
432 ms of 100 ms
Exceeds target by 332 ms
RAM
98 kB of 1024 kB
ROM
81 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS)
CLASSIFICATION (KERAS)

0.0005 | 20 | 84%

Type Filters Kernel Rate
conv2d 8 3 -
conv2d 5 3 -
dropout - - 0.25
dense 8 - -

9/28/2022, 8:00:22 AM

84%
grayscale-conv2d-b6a
PERFORMANCE
LATENCY
12 ms of 100 ms
RAM
98 kB of 1024 kB
ROM
103 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS)
CLASSIFICATION (KERAS)

0.0005 | 20 | 84%

Type Filters Kernel Rate
conv2d 8 3 -
conv2d 5 3 -
dropout - - 0.25
dense 16 - -

9/28/2022, 8:11:13 AM

82%
grayscale-conv2d-d62
PERFORMANCE
LATENCY
252 ms of 100 ms
Exceeds target by 152 ms
RAM
75 kB of 1024 kB
ROM
65 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS)
CLASSIFICATION (KERAS)

0.0005 | 20 | 82%

Type Filters Kernel Rate
conv2d 6 3 -
conv2d 3 3 -
dropout - - 0.25
dense 4 - -

9/28/2022, 8:08:17 AM

81%
grayscale-conv2d-11b
PERFORMANCE
LATENCY
322 ms of 100 ms
Exceeds target by 222 ms
RAM
75 kB of 1024 kB
ROM
68 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS)
CLASSIFICATION (KERAS)

0.0005 | 20 | 81%

Type Filters Kernel Rate
conv2d 6 3 -
conv2d 3 3 -
dropout - - 0.25
dense 6 - -

9/28/2022, 8:00:34 AM

78%
grayscale-conv2d-236
PERFORMANCE
LATENCY
13 ms of 100 ms
RAM
98 kB of 1024 kB
ROM
69 kB of 2048 kB
DSP NN Unused
IMAGE INPUT

96 |
96

IMAGE

Grayscale

ACCURACY (KERAS)
CLASSIFICATION (KERAS)

0.0005 | 20 | 78%

Type Filters Kernel Rate
conv2d 8 3 -
conv2d 5 3 -
dropout - - 0.25
dense 4 - -

9/28/2022, 8:00:28 AM

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