Hugging Face's logo

STMicroelectronics
/
fdmobilenet_pt

Image Classification
Model card Files
xet
 Community
FBAGSTM commited on
Commit
e470512
·
verified ·
1 Parent(s): b989a92

Release AI-ModelZoo-4.0.0

Browse files
Files changed (1) hide show
  1. README.md +117 -3
README.md CHANGED
@@ -1,3 +1,117 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ pipeline_tag: image-classification
4
+ ---
5
+ # FDMobileNet
6
+
7
+ ## **Use case** : `Image classification`
8
+
9
+ # Model description
10
+
11
+
12
+ Fast-Downsampling MobileNet (FDMobileNet) is an optimized variant of MobileNet designed for **extremely fast inference**. It achieves speed improvements through aggressive early spatial reduction while maintaining reasonable accuracy.
13
+
14
+ FDMobileNet employs a **fast downsampling strategy** that reduces spatial dimensions early in the network to minimize computation. It retains **depthwise separable convolutions** inherited from MobileNet for parameter efficiency, and uses a **width multiplier (Alpha)** to scale the number of channels (a025 = 0.25x, a050 = 0.5x, a075 = 0.75x).
15
+
16
+ Among the fastest models in the model zoo, FDMobileNet is ideal for ultra-low-latency real-time applications and battery-powered devices with strict power constraints.
17
+
18
+ (source: https://arxiv.org/abs/1802.03750)
19
+
20
+ The model is quantized to **int8** using **ONNX Runtime** and exported for efficient deployment.
21
+
22
+ ## Network information
23
+
24
+
25
+ | Network Information | Value |
26
+ |--------------------|-------|
27
+ | Framework | Torch |
28
+ | MParams | ~0.37–1.77 M |
29
+ | Quantization | Int8 |
30
+ | Provenance | https://github.com/qinzheng93/FD-MobileNet |
31
+ | Paper | https://arxiv.org/abs/1802.03750 |
32
+
33
+ ## Network inputs / outputs
34
+
35
+
36
+ For an image resolution of NxM and P classes
37
+
38
+ | Input Shape | Description |
39
+ | ----- | ----------- |
40
+ | (1, N, M, 3) | Single NxM RGB image with UINT8 values between 0 and 255 |
41
+
42
+ | Output Shape | Description |
43
+ | ----- | ----------- |
44
+ | (1, P) | Per-class confidence for P classes in FLOAT32|
45
+
46
+
47
+ ## Recommended platforms
48
+
49
+
50
+ | Platform | Supported | Recommended |
51
+ |----------|-----------|-----------|
52
+ | STM32L0 |[]|[]|
53
+ | STM32L4 |[]|[]|
54
+ | STM32U5 |[]|[]|
55
+ | STM32H7 |[]|[]|
56
+ | STM32MP1 |[]|[]|
57
+ | STM32MP2 |[]|[]|
58
+ | STM32N6 |[x]|[x]|
59
+
60
+ # Performances
61
+
62
+ ## Metrics
63
+
64
+ - Measures are done with default STEdgeAI Core configuration with enabled input / output allocated option.
65
+ - All the models are trained from scratch on Imagenet dataset
66
+
67
+ ### Reference **NPU** memory footprint on Imagenet dataset (see Accuracy for details on dataset)
68
+ | Model | Dataset | Format | Resolution | Series | Internal RAM (KiB) | External RAM (KiB) | Weights Flash (KiB) | STEdgeAI Core version |
69
+ |-------|---------|--------|------------|--------|--------------|--------------|---------------|----------------------|
70
+ | [fdmobilenet_a025_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6 | 294 | 0 | 377.03 | 3.0.0 |
71
+ | [fdmobilenet_a050_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6 | 343 | 0 | 973.39 | 3.0.0 |
72
+ | [fdmobilenet_a075_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6 | 441 | 0 | 1813.66 | 3.0.0 |
73
+
74
+
75
+
76
+ ### Reference **NPU** inference time on Imagenet dataset (see Accuracy for details on dataset)
77
+ | Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STEdgeAI Core version |
78
+ |-------|---------|--------|--------|------------|-------|-----------------|-------------------|---------------------|
79
+ | [fdmobilenet_a025_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6570-DK | NPU/MCU | 1.88 | 531.91 | 3.0.0 |
80
+ | [fdmobilenet_a050_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6570-DK | NPU/MCU | 4.07 | 245.70 | 3.0.0 |
81
+ | [fdmobilenet_a075_pt_224](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224_qdq_int8.onnx) | Imagenet | Int8 | 224×224×3 | STM32N6570-DK | NPU/MCU | 6.83 | 146.41 | 3.0.0 |
82
+
83
+
84
+ ### Accuracy with Imagenet dataset
85
+
86
+ | Model | Format | Resolution | Top 1 Accuracy |
87
+ | --- | --- | --- | --- |
88
+ | [fdmobilenet_a025_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224.onnx) | Float | 224x224x3 | 45.37 % |
89
+ | [fdmobilenet_a025_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 29.73 % |
90
+ | [fdmobilenet_a050_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224.onnx) | Float | 224x224x3 | 58.04 % |
91
+ | [fdmobilenet_a050_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 41.58 % |
92
+ | [fdmobilenet_a075_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224.onnx) | Float | 224x224x3 | 62.10 % |
93
+ | [fdmobilenet_a075_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 60.29 % |
94
+
95
+
96
+ | Model | Format | Resolution | Top 1 Accuracy |
97
+ | --- | --- | --- | --- |
98
+ | [fdmobilenet_a025_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224.onnx) | Float | 224x224x3 | 45.37 % |
99
+ | [fdmobilenet_a025_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a025_pt_224/fdmobilenet_a025_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 29.73 % |
100
+ | [fdmobilenet_a050_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224.onnx) | Float | 224x224x3 | 58.04 % |
101
+ | [fdmobilenet_a050_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a050_pt_224/fdmobilenet_a050_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 41.58 % |
102
+ | [fdmobilenet_a075_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224.onnx) | Float | 224x224x3 | 62.10 % |
103
+ | [fdmobilenet_a075_pt](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/fdmobilenet_pt/Public_pretrainedmodel_public_dataset/Imagenet/fdmobilenet_a075_pt_224/fdmobilenet_a075_pt_224_qdq_int8.onnx) | Int8 | 224x224x3 | 60.29 % |
104
+
105
+
106
+
107
+ ## Retraining and Integration in a simple example:
108
+
109
+ Please refer to the stm32ai-modelzoo-services GitHub [here](https://github.com/STMicroelectronics/stm32ai-modelzoo-services)
110
+
111
+
112
+
113
+ # References
114
+
115
+ <a id="1">[1]</a> - **Dataset**: Imagenet (ILSVRC 2012) — https://www.image-net.org/
116
+
117
+ <a id="2">[2]</a> - **Model**: FD-MobileNet — https://arxiv.org/abs/1802.03750