Image Classification
FBAGSTM commited on
Commit
6ee346b
·
verified ·
1 Parent(s): 0b65696

Release AI-ModelZoo-4.0.0

Browse files
Files changed (1) hide show
  1. README.md +95 -113
README.md CHANGED
@@ -71,74 +71,79 @@ For an image resolution of NxM and P classes
71
  - `tl` 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.
72
  - `fft` stands for "full fine-tuning", meaning that the full model weights were initialized from a transfer learning pre-trained model, and all the layers were unfrozen during the training.
73
 
74
- ### Reference **NPU** memory footprint on food-101 and ImageNet dataset (see Accuracy for details on dataset)
75
- |Model | Dataset | Format | Resolution | Series | Internal RAM | External RAM | Weights Flash | STM32Cube.AI version | STEdgeAI Core version |
76
- |----------|------------------|--------|-------------|------------------|------------------|---------------------|---------------|----------------------|-------------------------|
77
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 304.72 | 10.2.0 | 2.2.0 |
78
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 992.67 | 10.2.0 | 2.2.0 |
79
- | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_1.0_224_fft/mobilenet_v1_1.0_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6 | 1568 | 0.0 | 3602.97 | 10.2.0 | 2.2.0 |
80
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224_int8.tflite) | ImageNet | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 533.38 | 10.2.0 | 2.2.0 |
81
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224_int8.tflite) | ImageNet | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 1446.06 | 10.2.0 | 2.2.0 |
82
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224_int8.tflite) | ImageNet | Int8 | 224x224x3 | STM32N6 | 1568 | 0.0 | 4505.86 | 10.2.0 | 2.2.0 |
83
-
84
- ### Reference **NPU** inference time on food-101 and ImageNet dataset (see Accuracy for details on dataset)
85
-
86
-
87
- | Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STM32Cube.AI version | STEdgeAI Core version |
88
- |--------|----------|--------|-------------|------------------|------------------|---------------------|-----------|----------------------|-------------------------|
89
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 2.81 | 355.87 | 10.2.0 | 2.2.0 |
90
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 6.03 | 165.83 | 10.2.0 | 2.2.0 |
91
- | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_1.0_224_fft/mobilenet_v1_1.0_224_fft_int8.tflite) | food-101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 16.79 | 59.55 | 10.2.0 | 2.2.0 |
92
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 3.56 | 280.89 | 10.2.0 | 2.2.0 |
93
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 7.35 | 136.05 | 10.2.0 | 2.2.0 |
94
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 19.26 | 51.93 | 10.2.0 | 2.2.0 |
95
-
96
-
97
- ### Reference **MCU** memory footprint based on Flowers dataset and ImageNet dataset (see Accuracy for details on dataset)
98
-
99
- | Model | Format | Resolution | Series | Activation RAM | Runtime RAM | Weights Flash | Code Flash | Total RAM | Total Flash | STM32Cube.AI version |
100
- |--------------------------------------------------------------------------------------------------------------------------------------|--------|------------|---------|----------------|-------------|---------------|------------|-------------|-------------|-----------------------|
101
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 272.96 KiB | 16.38 KiB | 214.69 KiB | 67.24 KiB | 289.34 KiB | 281.93 KiB | 10.2.0 |
102
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 449.58 KiB | 16.38 KiB | 812.61 KiB | 80.61 KiB | 465.96 KiB | 893.22 KiB | 10.2.0 |
103
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | STM32H7 | 66.96 KiB | 16.33 KiB | 214.69 KiB | 67.19 KiB | 83.29 KiB | 281.88 KiB | 10.2.0 |
104
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | STM32H7 | 52.8 KiB | 16.33 KiB | 214.55 KiB | 69.28 KiB | 69.13 KiB | 283.83 KiB | 10.2.0 |
105
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 267.2 KiB | 16.44 KiB | 467.33 KiB | 68.37 KiB | 283.64 KiB | 535.7 KiB | 10.2.0 |
106
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 404.28 KiB | 16.44 KiB | 1314 KiB | 81.72 KiB | 447.51 KiB | 1395.72 KiB | 10.2.0 |
107
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 1331.13 KiB | 16.48 KiB | 4157.09 KiB | 108.46 KiB | 1347.61 KiB | 4265.55 KiB | 10.2.0 |
108
-
109
-
110
- ### Reference **MCU** inference time based on Flowers dataset and ImageNet dataset (see Accuracy for details on dataset)
111
-
112
-
113
- | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
114
- |-------------------|--------|------------|------------------|------------------|-----------|---------------------|-----------------------|
115
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 166.9 ms | 10.2.0 |
116
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 471.68 ms | 10.2.0 |
117
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 30.63 ms | 10.2.0 |
118
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | STM32H747I-DISCO | 1 CPU | 400 MHz | 29.04 ms | 10.2.0 |
119
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 170.37 ms | 10.2.0 |
120
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 477.79 ms | 10.2.0 |
121
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 1656.41 ms | 10.2.0 |
 
 
122
 
123
 
124
  ### Reference **MPU** inference time based on Flowers dataset (see Accuracy for details on dataset)
125
- | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
126
- |-----------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
127
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 14.27 ms | 7.54 | 92.46 | 0 | v6.1.0 | OpenVX |
128
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 32.79 ms | 3.83 | 96.17 | 0 | v6.1.0 | OpenVX |
129
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 3.81 ms | 15.36 | 84.64 | 0 | v6.1.0 | OpenVX |
130
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 3.66 ms | 13.91 | 86.09 | 0 | v6.1.0 | OpenVX |
131
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 33.91ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
132
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 90.6 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
133
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 6.32 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
134
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 5.83 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
135
- |[MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 52.39 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
136
- |[MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 144.47 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
137
- |[MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 9.31 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
138
- |[MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 9.37 ms | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
 
139
 
140
  ** **To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization**
141
 
 
 
142
  ### Accuracy with Flowers dataset
143
 
144
 
@@ -146,22 +151,14 @@ Dataset details: [link](http://download.tensorflow.org/example_images/flower_pho
146
 
147
  | Model | Format | Resolution | Top 1 Accuracy |
148
  |-------|--------|------------|----------------|
149
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs.h5) | Float | 224x224x3 | 88.83 % |
150
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs_int8.tflite) | Int8 | 224x224x3 | 89.37 % |
151
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl.h5) | Float | 224x224x3 | 85.83 % |
152
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl_int8.tflite) | Int8 | 224x224x3 | 83.24 % |
153
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft.h5) | Float | 224x224x3 | 93.05 % |
154
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | 92.1 % |
155
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs.h5) | Float | 224x224x3 | 92.1 % |
156
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs_int8.tflite) | Int8 | 224x224x3 | 91.55 % |
157
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl.h5) | Float | 224x224x3 | 88.56 % |
158
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl_int8.tflite) | Int8 | 224x224x3 | 87.74 % |
159
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft.h5) | Float | 224x224x3 | 95.1 % |
160
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | 94.41 % |
161
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft.h5) | Float | 96x96x3 | 87.47 % |
162
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_fft/mobilenet_v1_0.25_96_fft_int8.tflite) | Int8 | 96x96x3 | 87.06 % |
163
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs.h5) | Float | 96x96x1 | 74.93 % |
164
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/flowers/mobilenet_v1_0.25_96_grayscale_tfs/mobilenet_v1_0.25_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | 74.93 % |
165
 
166
 
167
 
@@ -172,18 +169,10 @@ Dataset details: [link](https://data.mendeley.com/datasets/tywbtsjrjv/1) , Licen
172
 
173
  | Model | Format | Resolution | Top 1 Accuracy |
174
  |-------|--------|------------|----------------|
175
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs.h5) | Float | 224x224x3 | 99.92 % |
176
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs_int8.tflite) | Int8 | 224x224x3 | 99.92 % |
177
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl.h5) | Float | 224x224x3 | 85.38 % |
178
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl_int8.tflite) | Int8 | 224x224x3 | 83.7 % |
179
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft.h5) | Float | 224x224x3 | 99.95 % |
180
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | 99.82 % |
181
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs.h5) | Float | 224x224x3 | 99.9 % |
182
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs_int8.tflite) | Int8 | 224x224x3 | 99.83 % |
183
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl.h5) | Float | 224x224x3 | 93.05 % |
184
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl_int8.tflite) | Int8 | 224x224x3 | 92.7 % |
185
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft.h5) | Float | 224x224x3 | 99.94 % |
186
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant-village/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | 99.85 % |
187
 
188
 
189
  ### Accuracy with Food-101 dataset
@@ -193,23 +182,15 @@ Dataset details: [link](https://data.vision.ee.ethz.ch/cvl/datasets_extra/food-1
193
 
194
  | Model | Format | Resolution | Top 1 Accuracy |
195
  |-------|--------|------------|----------------|
196
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs.h5) | Float | 224x224x3 | 72.16 % |
197
- | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_tfs/mobilenet_v1_0.25_224_tfs_int8.tflite) | Int8 | 224x224x3 | 71.13 % |
198
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl.h5) | Float | 224x224x3 | 43.21 % |
199
- | [MobileNet v1 0.25 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_tl/mobilenet_v1_0.25_224_tl_int8.tflite) | Int8 | 224x224x3 | 39.89 % |
200
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft.h5) | Float | 224x224x3 | 72.36 % |
201
- | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.25_224_fft/mobilenet_v1_0.25_224_fft_int8.tflite) | Int8 | 224x224x3 | 69.52 % |
202
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs.h5) | Float | 224x224x3 | 76.97 % |
203
- | [MobileNet v1 0.5 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_tfs/mobilenet_v1_0.5_224_tfs_int8.tflite) | Int8 | 224x224x3 | 76.37 % |
204
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl.h5) | Float | 224x224x3 | 48.78 % |
205
- | [MobileNet v1 0.5 tl](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_tl/mobilenet_v1_0.5_224_tl_int8.tflite) | Int8 | 224x224x3 | 45.89 % |
206
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft.h5) | Float | 224x224x3 | 76.72 % |
207
- | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_0.5_224_fft/mobilenet_v1_0.5_224_fft_int8.tflite) | Int8 | 224x224x3 | 74.82 % |
208
- | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_1.0_224_fft/mobilenet_v1_1.0_224_fft.h5) | Float | 224x224x3 | 80.38 % |
209
- | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food-101/mobilenet_v1_1.0_224_fft/mobilenet_v1_1.0_224_fft_int8.tflite) | Int8 | 224x224x3 | 79.43 % |
210
-
211
-
212
- ### Accuracy with ImageNet dataset
213
 
214
  Dataset details: [link](https://www.image-net.org), Quotation[[4]](#4).
215
  Number of classes: 1000.
@@ -218,12 +199,13 @@ For the sake of simplicity, the accuracy reported here was estimated on the 5000
218
 
219
  |model | Format | Resolution | Top 1 Accuracy |
220
  |---------|--------|------------|----------------|
221
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224.h5) | Float | 224x224x3 | 48.96 % |
222
- | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.25_224/mobilenet_v1_0.25_224_int8.tflite) | Int8 | 224x224x3 | 46.34 % |
223
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224.h5) | Float | 224x224x3 | 62.11 % |
224
- | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_0.5_224/mobilenet_v1_0.5_224_int8.tflite) | Int8 | 224x224x3 | 59.92 % |
225
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224.h5) | Float | 224x224x3 | 69.52 % |
226
- | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/ImageNet/mobilenet_v1_1.0_224/mobilenet_v1_1.0_224_int8.tflite) | Int8 | 224x224x3 | 68.64 % |
 
227
 
228
 
229
 
 
71
  - `tl` 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.
72
  - `fft` stands for "full fine-tuning", meaning that the full model weights were initialized from a transfer learning pre-trained model, and all the layers were unfrozen during the training.
73
 
74
+ ### Reference **NPU** memory footprint on food101 and imagenet dataset (see Accuracy for details on dataset)
75
+ |Model | Dataset | Format | Resolution | Series | Internal RAM | External RAM | Weights Flash | STEdgeAI Core version |
76
+ |----------|------------------|--------|-------------|------------------|------------------|---------------------|---------------|-------------------------|
77
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6 | 392 | 0.0 | 240.88 | 3.0.0 |
78
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 864.99 | 3.0.0 |
79
+ | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a100_224_fft/mobilenetv1_a100_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6 | 1568 | 0.0 | 3347.59 | 3.0.0 |
80
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224_int8.tflite) | imagenet | Int8 | 224x224x3 | STM32N6 | 392 | 0.0 | 469.53 | 3.0.0 |
81
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_int8.tflite) | imagenet | Int8 | 224x224x3 | STM32N6 | 588 | 0.0 | 1318.38 | 3.0.0 |
82
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_qdq_w4_38.8%_w8_61.2%_a8_100%_acc_60.87.onnx) | imagenet | Int8/Int4 | 224x224x3 | STM32N6 | 588 | 0.0 | 1067.95 | 3.0.0 |
83
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224_int8.tflite) | imagenet | Int8 | 224x224x3 | STM32N6 | 1568 | 0.0 | 4250.49 | 3.0.0 |
84
+
85
+ ### Reference **NPU** inference time on food101 and imagenet dataset (see Accuracy for details on dataset)
86
+
87
+
88
+ | Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STEdgeAI Core version |
89
+ |--------|----------|--------|-------------|------------------|------------------|---------------------|-----------|-------------------------|
90
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 2.37 | 421.94 | 3.0.0 |
91
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 5.38 | 185.87 | 3.0.0 |
92
+ | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a100_224_fft/mobilenetv1_a100_224_fft_int8.tflite) | food101 | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 16.36 | 61.12 | 3.0.0 |
93
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 3.1 | 322.58 | 3.0.0 |
94
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 6.67 | 149.92 | 3.0.0 |
95
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_qdq_w4_38.8%_w8_61.2%_a8_100%_acc_60.87.onnx) | Imagenet | Int8/Int4 | 224x224x3 | STM32N6570-DK | NPU/MCU | 5.95 | 168.07 | 3.0.0 |
96
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 18.81 | 53.16 | 3.0.0 |
97
+
98
+
99
+ ### Reference **MCU** memory footprint based on Flowers dataset and imagenet dataset (see Accuracy for details on dataset)
100
+
101
+ | Model | Dataset| Format | Resolution | Series | Activation RAM | Runtime RAM | Weights Flash | Code Flash | Total RAM | Total Flash | STEdgeAI Core version |
102
+ |-------------|--------|--------|------------|---------|----------------|---------------|------------|-------------|-------------|-------------|------------------------|
103
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) |tf_flowers| Int8 | 224x224x3 | STM32H7 | 271.04 KiB | 0.7 KiB | 214.69 KiB | 36.07 KiB | 271.74 KiB | 250.76 KiB | 3.0.0 |
104
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | tf_flowers| Int8 | 224x224x3 | STM32H7 | 456.67 KiB | 0.7 KiB | 812.61 KiB | 46.79 KiB | 457.37 KiB | 859.4 KiB | 3.0.0 |
105
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | tf_flowers | Int8 | 96x96x3 | STM32H7 | 63.04 KiB | 0.7 KiB | 214.69 KiB | 36.03 KiB | 63.74 KiB | 250.72 KiB | 3.0.0 |
106
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | tf_flowers| Int8 | 96x96x1 | STM32H7 | 52.8 KiB | 0.3 KiB | 214.55 KiB | 39.13 KiB | 53.1 KiB | 253.68 KiB | 3.0.0 |
107
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32H7 | 267.2 KiB | 0.3 KiB | 467.33 KiB | 37.61 KiB | 267.5 KiB | 504.94 KiB | 3.0.0 |
108
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32H7 | 431.07 KiB | 0.3 KiB | 1314 KiB | 48.32 KiB | 431.37 KiB | 1362.32 KiB | 3.0.0 |
109
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224_int8.tflite) | Imagenet | Int8 | 224x224x3 | STM32H7 | 899.78 KiB | 0.3 KiB | 4157.09 KiB | 69.82 KiB | 900.08 KiB | 4226.91 KiB | 3.0.0 |
110
+
111
+
112
+ ### Reference **MCU** inference time based on Flowers dataset and imagenet dataset (see Accuracy for details on dataset)
113
+
114
+
115
+ | Model | Dataset | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STEdgeAI Core version |
116
+ |------------|------|--------|------------|------------------|------------------|-----------|---------------------|------------------------|
117
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | tf_flowers | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 176.26 ms | 3.0.0 |
118
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | tf_flowers | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 488.31 ms | 3.0.0 |
119
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | tf_flowers| Int8 | 96x96x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 32.64 ms | 3.0.0 |
120
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | tf_flowers| Int8 | 96x96x1 | STM32H747I-DISCO | 1 CPU | 400 MHz | 29.62 ms | 3.0.0 |
121
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224_int8.tflite) | Imagenet| Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 180.87 ms | 3.0.0 |
122
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_int8.tflite) | Imagenet| Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 504.08 ms | 3.0.0 |
123
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224_int8.tflite) | Imagenet| Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 1651.05 ms | 3.0.0 |
124
 
125
 
126
  ### Reference **MPU** inference time based on Flowers dataset (see Accuracy for details on dataset)
127
+
128
+ | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
129
+ |-----------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|------|-------|------|--------------------|-----------------------|
130
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 14.06 | 7.47 | 92.53 | 0 | v6.1.0 | OpenVX |
131
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 32.37 | 3.84 | 96.16 | 0 | v6.1.0 | OpenVX |
132
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 3.62 | 18.33| 81.67 | 0 | v6.1.0 | OpenVX |
133
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 3.72 | 14.97| 85.03 | 0 | v6.1.0 | OpenVX |
134
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 31.70 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
135
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 89.23 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
136
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 5.99 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
137
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 5.94 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
138
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 49.86 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
139
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 142.62 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
140
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | Int8 | 96x96x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 9.18 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
141
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 9.24 | NA | NA | 100 | v6.1.0 | TensorFlowLite 2.18.0 |
142
 
143
  ** **To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization**
144
 
145
+ ** **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.
146
+
147
  ### Accuracy with Flowers dataset
148
 
149
 
 
151
 
152
  | Model | Format | Resolution | Top 1 Accuracy |
153
  |-------|--------|------------|----------------|
154
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft.keras) | Float | 224x224x3 | 93.05 % |
155
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | 92.1 % |
156
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft.keras) | Float | 224x224x3 | 95.1 % |
157
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | 94.41 % |
158
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft.keras) | Float | 96x96x3 | 87.47 % |
159
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_fft/mobilenetv1_a025_96_fft_int8.tflite) | Int8 | 96x96x3 | 87.06 % |
160
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs.keras) | Float | 96x96x1 | 74.93 % |
161
+ | [MobileNet v1 0.25 tfs](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/tf_flowers/mobilenetv1_a025_96_grayscale_tfs/mobilenetv1_a025_96_grayscale_tfs_int8.tflite) | Int8 | 96x96x1 | 74.93 % |
 
 
 
 
 
 
 
 
162
 
163
 
164
 
 
169
 
170
  | Model | Format | Resolution | Top 1 Accuracy |
171
  |-------|--------|------------|----------------|
172
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant_leaf_diseases/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft.keras) | Float | 224x224x3 | 99.95 % |
173
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant_leaf_diseases/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | 99.82 % |
174
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant_leaf_diseases/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft.keras) | Float | 224x224x3 | 99.94 % |
175
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/plant_leaf_diseases/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | 99.85 % |
 
 
 
 
 
 
 
 
176
 
177
 
178
  ### Accuracy with Food-101 dataset
 
182
 
183
  | Model | Format | Resolution | Top 1 Accuracy |
184
  |-------|--------|------------|----------------|
185
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft.keras) | Float | 224x224x3 | 75.75 % |
186
+ | [MobileNet v1 0.25 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a025_224_fft/mobilenetv1_a025_224_fft_int8.tflite) | Int8 | 224x224x3 | 73.24 % |
187
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft.keras) | Float | 224x224x3 | 82.06 % |
188
+ | [MobileNet v1 0.5 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a050_224_fft/mobilenetv1_a050_224_fft_int8.tflite) | Int8 | 224x224x3 | 80.64 % |
189
+ | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a100_224_fft/mobilenetv1_a100_224_fft.keras) | Float | 224x224x3 | 84.57 % |
190
+ | [MobileNet v1 1.0 fft](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/ST_pretrainedmodel_public_dataset/food101/mobilenetv1_a100_224_fft/mobilenetv1_a100_224_fft_int8.tflite) | Int8 | 224x224x3 | 83.07 % |
191
+
192
+
193
+ ### Accuracy with imagenet dataset
 
 
 
 
 
 
 
 
194
 
195
  Dataset details: [link](https://www.image-net.org), Quotation[[4]](#4).
196
  Number of classes: 1000.
 
199
 
200
  |model | Format | Resolution | Top 1 Accuracy |
201
  |---------|--------|------------|----------------|
202
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224.keras) | Float | 224x224x3 | 50.5 % |
203
+ | [MobileNet v1 0.25](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a025_224/mobilenetv1_a025_224_int8.tflite) | Int8 | 224x224x3 | 47.94 % |
204
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224.keras) | Float | 224x224x3 | 64.02 % |
205
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_int8.tflite) | Int8 | 224x224x3 | 62.25 % |
206
+ | [MobileNet v1 0.5](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a050_224/mobilenetv1_a050_224_qdq_w4_38.8%_w8_61.2%_a8_100%_acc_60.87.onnx) | Int8/Int4 | 224x224x3 | 60.87 % |
207
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224.keras) | Float | 224x224x3 | 70.92 % |
208
+ | [MobileNet v1 1.0](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/mobilenetv1/Public_pretrainedmodel_public_dataset/imagenet/mobilenetv1_a100_224/mobilenetv1_a100_224_int8.tflite) | Int8 | 224x224x3 | 69.64 % |
209
 
210
 
211