Benchmark - IDS NXT rome Technical Manual

Filter kernel/pooling parameters
Inference time

11.1.2 Benchmark

Reference models from tensorflow.keras.applications
Architecture
MobileNet V1 α=1.0
MobileNet V1 α=0.75
MobileNet V1 α=0.5
MobileNet V1 α=0.5
MobileNet V1 α=0.25
MobileNet V1 α=0.25
MobileNet V2 α=1.4
MobileNet V2 α=1.0
MobileNet V2 α=0.5
MobileNet V3 large alpha 1.0
MobileNet V3 large alpha 0.75
MobileNet V3 small alpha 1.0
ResNet50
Xception
MobileNet_V1_SSD
2021-11-08
·
Squeeze-and-Excitation layer
·
ReLU activation
·
ReLU6 activation
·
Swish activation
·
Sigmoid activation
·
Batch normalization
·
Kernel/pooling window: any rectangle up to 15 x 15 pixels
·
Kernel depth: any depth
·
Kernel/pooling stride: up to 15 pixels
See Benchmark
Input format output format
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(128, 128, 3) > (1000)
(224, 224, 3) > (1000)
(128, 128, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(224, 224, 3) > (1000)
(300, 300, 3) > (scores: (3323, 81),
boxes: (3323, 4), anchors: (3323, 4))
IDS NXT: Technical manual IDS NXT rome
Single image inference time
[ms]
66
50
34
16
28
12
102
71
47
73
65
33
297
596
132
31