Hugging Face's logo

STMicroelectronics
/
fdmobilenet

Image Classification
Model card Files
xet
 Community
FBAGSTM commited on
Commit
e4072fe
·
verified ·
1 Parent(s): a030392

Release AI-ModelZoo-4.0.0

Browse files
Files changed (1) hide show
  1. README.md +54 -51
README.md CHANGED
@@ -64,64 +64,67 @@ For an image resolution of NxM and P classes and 0.25 alpha parameter :
64
  ## Metrics
65
 
66
  * Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
67
- * `tfs` stands for "transfer learning", meaning that the model backbone weights were initialized from a pre-trained model, then only the last layer was unfrozen during the training.
68
 
69
- ### Reference **NPU** memory footprint on food-101 dataset (see Accuracy for details on dataset)
70
- |Model | Format | Resolution | Series | Internal RAM (KiB)| External RAM (KiB)| Weights Flash (KiB) | STM32Cube.AI version | STEdgeAI Core version |
71
- |----------|--------|-------------|------------------|------------------|---------------------|---------------------|----------------------|-------------------------|
72
- | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/fdmobilenet_0.25_224_tfs/fdmobilenet_0.25_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6 | 294 |0.0| 197.56 | 10.2.0 | 2.2.0 |
73
- | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/st_fdmobilenet_v1_224_tfs/st_fdmobilenet_v1_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6 | 294 | 0.0 | 222.33 | 10.2.0 | 2.2.0 |
74
- | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/fdmobilenet_0.25_128_tfs/fdmobilenet_0.25_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6 | 96 | 0.0 | 197.56 | 10.2.0 | 2.2.0 |
75
- | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/st_fdmobilenet_v1_128_tfs/st_fdmobilenet_v1_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6 | 96 | 0.0 | 222.33 | 10.2.0 | 2.2.0 |
76
 
77
 
78
- ### Reference **NPU** inference time on food-101 dataset (see Accuracy for details on dataset)
79
- | Model | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STM32Cube.AI version | STEdgeAI Core version |
80
- |--------|--------|-------------|------------------|------------------|---------------------|-------|----------------------|-------------------------|
81
- | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/fdmobilenet_0.25_224_tfs/fdmobilenet_0.25_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 1.46 | 684.93 | 10.2.0 | 2.2.0 |
82
- | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/st_fdmobilenet_v1_224_tfs/st_fdmobilenet_v1_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 1.81 | 552.49 | 10.2.0 | 2.2.0 |
83
- | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/fdmobilenet_0.25_128_tfs/fdmobilenet_0.25_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6570-DK | NPU/MCU | 0.93 | 1075.27 | 10.2.0 | 2.2.0 |
84
- | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food-101/st_fdmobilenet_v1_128_tfs/st_fdmobilenet_v1_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6570-DK | NPU/MCU | 1.07 | 934.58 | 10.2.0 | 2.2.0 |
85
 
86
 
87
  ### Reference **MCU** memory footprints based on Flowers dataset (see Accuracy for details on dataset)
88
- | Model | Format | Resolution | Series | Activation RAM | Runtime RAM | Weights Flash | Code Flash | Total RAM | Total Flash | STM32Cube.AI version |
89
- |-----------------------|--------|--------------|---------|----------------|-------------|---------------|------------|------------|-------------|----------------------|
90
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | STM32H7 | 157.03 KiB | 14.25 KiB | 128.32 KiB | 57.01 KiB | 171.28 KiB | 185.33 KiB | 10.2.0 |
91
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32H7 | 211.64 KiB | 14.25 KiB | 144.93 KiB | 58.51 KiB | 225.89 KiB | 203.44 KiB | 10.2.0 |
92
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | STM32H7 | 56.16 KiB | 14.2 KiB | 128.32 KiB | 56.98 KiB | 70.36 KiB | 185.3 KiB | 10.2.0 |
93
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32H7 | 74.23 KiB | 14.2 KiB | 144.93 KiB | 58.47 KiB | 88.43 KiB | 203.4 KiB | 10.2.0 |
94
 
95
 
96
  ### Reference **MCU** inference time based on Flowers dataset (see Accuracy for details on dataset)
97
- | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
98
- |-----------------------|--------|--------------|------------------|------------------|---------------|---------------------|----------------------|
99
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 54.36 ms | 10.2.0 |
100
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 103.45 ms | 10.2.0 |
101
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 18.04 ms | 10.2.0 |
102
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 32.69 ms | 10.2.0 |
103
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32F769I-DISCO | 1 CPU | 216 MHz | 176.5 ms | 10.2.0 |
104
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32F769I-DISCO | 1 CPU | 216 MHz | 59.29 ms | 10.2.0 |
105
 
106
 
107
  ### Reference **MPU** inference time based on Flowers dataset (see Accuracy for details on dataset)
108
- | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
109
- |-----------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
110
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 6.59 ms | 13.71 | 86.29 | 0 | v6.1.0 | OpenVX |
111
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 7.79 ms | 11.59 | 88.41 | 0 | v6.1.0 | OpenVX |
112
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 2.23 ms | 19.22 | 80.78 | 0 | v6.1.0 | OpenVX |
113
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 2.84 ms | 18.90 | 81.10 | 0 | v6.1.0 | OpenVX |
114
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 22.87 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
115
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 39.05 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
116
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 7.98 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
117
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 13.54 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
118
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 33.72 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
119
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 60.14 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
120
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 10.88 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
121
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 19.59 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
 
122
 
123
  ** **To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization**
124
 
 
 
125
  ### Accuracy with Flowers dataset
126
  Dataset details: http://download.tensorflow.org/example_images/flower_photos.tgz , License CC - BY 2.0
127
  Number of classes: 5, 3670 files
@@ -160,14 +163,14 @@ Number of classes: 101, number of files: 101000
160
 
161
  | Model | Format | Resolution | Top 1 Accuracy (%) |
162
  |-----------------------|--------|--------------|----------------------|
163
- | FdMobileNet 0.25 tfs | Float | 224x224x3 | 60.41 |
164
- | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | 58.78 |
165
- | ST FdMobileNet v1 tfs | Float | 224x224x3 | 66.19 |
166
- | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | 64.71 |
167
- | FdMobileNet 0.25 tfs | Float | 128x128x3 | 45.58 |
168
- | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | 44.86 |
169
- | ST FdMobileNet v1 tfs | Float | 128x128x3 | 54.22 |
170
- | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | 53.74 |
171
 
172
 
173
  ## Retraining and Integration in a simple example:
 
64
  ## Metrics
65
 
66
  * Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
67
+ * `tfs` stands for "training from scratch", meaning that the model weights are randomly initialized before the training and all layers are actually trained.
68
 
69
+ ### Reference **NPU** memory footprint on food101 dataset (see Accuracy for details on dataset)
70
+ |Model | Format | Resolution | Series | Internal RAM (KiB)| External RAM (KiB)| Weights Flash (KiB) | STEdgeAI Core version |
71
+ |----------|--------|-------------|------------------|------------------|---------------------|---------------------|----------------------|
72
+ | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/fdmobilenet_a025_224_tfs/fdmobilenet_a025_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6 | 294 |0.0| 148.34 | 3.0.0 |
73
+ | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/st_fdmobilenetv1_224_tfs/st_fdmobilenetv1_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6 | 343 | 0.0 | 167.2 | 3.0.0 |
74
+ | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/fdmobilenet_a025_128_tfs/fdmobilenet_a025_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6 | 96 | 0.0 | 146.66 | 3.0.0 |
75
+ | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/st_fdmobilenetv1_128_tfs/st_fdmobilenetv1_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6 | 112 | 0.0 | 163.83 | 3.0.0 |
76
 
77
 
78
+ ### Reference **NPU** inference time on food101 dataset (see Accuracy for details on dataset)
79
+ | Model | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STEdgeAI Core version |
80
+ |--------|--------|-------------|------------------|------------------|---------------------|-------|----------------------|
81
+ | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/fdmobilenet_a025_224_tfs/fdmobilenet_a025_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 1.29 | 775.19 | 3.0.0 |
82
+ | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/st_fdmobilenetv1_224_tfs/st_fdmobilenetv1_224_tfs_int8.tflite) | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 1.67 | 598.8 | 3.0.0 |
83
+ | [FdMobileNet 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/fdmobilenet_a025_128_tfs/fdmobilenet_a025_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6570-DK | NPU/MCU | 0.75 | 1333.33 | 3.0.0 |
84
+ | [ST FdMobileNet v1 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet/ST_pretrainedmodel_public_dataset/food101/st_fdmobilenetv1_128_tfs/st_fdmobilenetv1_128_tfs_int8.tflite) | Int8 | 128x128x3 | STM32N6570-DK | NPU/MCU | 0.9 | 1111.11 | 3.0.0 |
85
 
86
 
87
  ### Reference **MCU** memory footprints based on Flowers dataset (see Accuracy for details on dataset)
88
+ | Model | Format | Resolution | Series | Activation RAM | Runtime RAM | Weights Flash | Code Flash | Total RAM | Total Flash | STEdgeAI Core version |
89
+ |-----------------------|--------|--------------|---------|----------------|-------------|---------------|------------|------------|-------------|-----------------------|
90
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | STM32H7 | 157.03 KiB | 0.3 KiB | 128.32 KiB | 29.99 KiB | 157.33 KiB | 158.31 KiB | 3.0.0 |
91
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32H7 | 211.64 KiB | 0.3 KiB | 144.93 KiB | 31.18 KiB | 211.94 KiB | 176.11 KiB | 3.0.0 |
92
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | STM32H7 | 56.16 KiB | 0.3 KiB | 128.32 KiB | 29.95 KiB | 56.46 KiB | 158.27 KiB | 3.0.0 |
93
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32H7 | 74.23 KiB | 0.3 KiB | 144.93 KiB | 31.13 KiB | 74.53 KiB | 176.06 KiB | 3.0.0 |
94
 
95
 
96
  ### Reference **MCU** inference time based on Flowers dataset (see Accuracy for details on dataset)
97
+ | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STEdgeAI Core version |
98
+ |-----------------------|--------|--------------|------------------|------------------|---------------|---------------------|-----------------------|
99
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 64.56 ms | 3.0.0 |
100
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 113.66 ms | 3.0.0 |
101
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 21.34 ms | 3.0.0 |
102
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 36.04 ms | 3.0.0 |
103
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | STM32F769I-DISCO | 1 CPU | 216 MHz | 176.5 ms | 3.0.0 |
104
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | STM32F769I-DISCO | 1 CPU | 216 MHz | 59.29 ms | 3.0.0 |
105
 
106
 
107
  ### Reference **MPU** inference time based on Flowers dataset (see Accuracy for details on dataset)
108
+
109
+ | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
110
+ |-----------------------|--------|------------|-----------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
111
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 6.48 | 15.25 | 84.75 | 0 | v6.1.0 | OpenVX |
112
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 7.82 | 16.29 | 83.71 | 0 | v6.1.0 | OpenVX |
113
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 2.01 | 18.25 | 81.75 | 0 | v6.1.0 | OpenVX |
114
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 2.78 | 13.80 | 86.20 | 0 | v6.1.0 | OpenVX |
115
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 24.96 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
116
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 43.05 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
117
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 8.96 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
118
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 14.19 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
119
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 34.86 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
120
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 63.78 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
121
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 11.86 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
122
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 20.34 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
123
 
124
  ** **To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization**
125
 
126
+ ** **Note:** On STM32MP2 devices, per-channel quantized models are internally converted to per-tensor quantization by the compiler using an entropy-based method. This may introduce a slight loss in accuracy compared to the original per-channel models.
127
+
128
  ### Accuracy with Flowers dataset
129
  Dataset details: http://download.tensorflow.org/example_images/flower_photos.tgz , License CC - BY 2.0
130
  Number of classes: 5, 3670 files
 
163
 
164
  | Model | Format | Resolution | Top 1 Accuracy (%) |
165
  |-----------------------|--------|--------------|----------------------|
166
+ | FdMobileNet 0.25 tfs | Float | 224x224x3 | 63.03 |
167
+ | FdMobileNet 0.25 tfs | Int8 | 224x224x3 | 62.11 |
168
+ | ST FdMobileNet v1 tfs | Float | 224x224x3 | 69.31 |
169
+ | ST FdMobileNet v1 tfs | Int8 | 224x224x3 | 68.73 |
170
+ | FdMobileNet 0.25 tfs | Float | 128x128x3 | 51.12 |
171
+ | FdMobileNet 0.25 tfs | Int8 | 128x128x3 | 50.26 |
172
+ | ST FdMobileNet v1 tfs | Float | 128x128x3 | 59.07 |
173
+ | ST FdMobileNet v1 tfs | Int8 | 128x128x3 | 58.15 |
174
 
175
 
176
  ## Retraining and Integration in a simple example: