Hugging Face's logo

qualcomm
/
MobileNet-v3-Small

Image Classification
PyTorch
backbone
bu_auto
real_time
android
Model card Files
xet
 Community
MobileNet-v3-Small
File size: 8,298 Bytes
2f853e0
 
e429491
2f853e0
 
e89d143
2f853e0
 
313f042
2f853e0
 
 
7ca11ec
2f853e0
c1f2411
fb50908
2f853e0
 
c1f2411
46d996e
c1f2411
 
 
 
 
 
 
 
 
 
 
 
2d632cd
 
 
 
c1f2411
 
 
 
 
 
46d996e
c1f2411
 
 
 
 
 
46d996e
c1f2411
 
 
 
 
 
 
 
 
 
 
 
 
 
2d632cd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cf5a84d
2f853e0
905439f
 
cf5a84d
2f853e0
 
 
 
 
97ec9cf
2f853e0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
---
library_name: pytorch
license: other
tags:
- backbone
- bu_auto
- real_time
- android
pipeline_tag: image-classification

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilenet_v3_small/web-assets/model_demo.png)

# MobileNet-v3-Small: Optimized for Qualcomm Devices

MobileNetV3Small is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This is based on the implementation of MobileNet-v3-Small found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/mobilenet_v3_small) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilenet_v3_small/releases/v0.51.0/mobilenet_v3_small-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilenet_v3_small/releases/v0.51.0/mobilenet_v3_small-qnn_dlc-float.zip)
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilenet_v3_small/releases/v0.51.0/mobilenet_v3_small-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/mobilenet_v3_small/releases/v0.51.0/mobilenet_v3_small-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[MobileNet-v3-Small on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/mobilenet_v3_small)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/mobilenet_v3_small) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [MobileNet-v3-Small on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/mobilenet_v3_small) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 2.54M
- Model size (float): 9.71 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| MobileNet-v3-Small | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.243 ms | 0 - 33 MB | NPU
| MobileNet-v3-Small | ONNX | float | Snapdragon® X2 Elite | 0.25 ms | 5 - 5 MB | NPU
| MobileNet-v3-Small | ONNX | float | Snapdragon® X Elite | 0.685 ms | 5 - 5 MB | NPU
| MobileNet-v3-Small | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.353 ms | 0 - 45 MB | NPU
| MobileNet-v3-Small | ONNX | float | Qualcomm® QCS8550 (Proxy) | 0.55 ms | 1 - 79 MB | NPU
| MobileNet-v3-Small | ONNX | float | Qualcomm® QCS9075 | 0.766 ms | 1 - 3 MB | NPU
| MobileNet-v3-Small | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.292 ms | 0 - 28 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.319 ms | 1 - 33 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Snapdragon® X2 Elite | 0.448 ms | 1 - 1 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Snapdragon® X Elite | 0.989 ms | 1 - 1 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.546 ms | 0 - 43 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 2.094 ms | 1 - 29 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 0.832 ms | 0 - 5 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® SA8775P | 1.088 ms | 1 - 31 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® QCS9075 | 0.983 ms | 1 - 3 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1.575 ms | 0 - 46 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® SA7255P | 2.094 ms | 1 - 29 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Qualcomm® SA8295P | 1.454 ms | 0 - 29 MB | NPU
| MobileNet-v3-Small | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.412 ms | 1 - 33 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.308 ms | 0 - 29 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.431 ms | 0 - 0 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® X Elite | 0.937 ms | 0 - 0 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.559 ms | 0 - 37 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 2.214 ms | 2 - 4 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 1.711 ms | 0 - 26 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 0.794 ms | 0 - 76 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® SA8775P | 0.99 ms | 0 - 27 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 0.952 ms | 0 - 2 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 2.801 ms | 0 - 144 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 0.982 ms | 0 - 40 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® SA7255P | 1.711 ms | 0 - 26 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Qualcomm® SA8295P | 1.302 ms | 0 - 24 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.372 ms | 0 - 29 MB | NPU
| MobileNet-v3-Small | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 0.799 ms | 0 - 25 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.319 ms | 0 - 35 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.558 ms | 0 - 45 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 2.168 ms | 0 - 30 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 0.833 ms | 0 - 10 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® SA8775P | 1.156 ms | 0 - 32 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® QCS9075 | 1.014 ms | 0 - 8 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1.596 ms | 0 - 46 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® SA7255P | 2.168 ms | 0 - 30 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Qualcomm® SA8295P | 1.49 ms | 0 - 29 MB | NPU
| MobileNet-v3-Small | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.424 ms | 0 - 30 MB | NPU

## License
* The license for the original implementation of MobileNet-v3-Small can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [Searching for MobileNetV3](https://arxiv.org/abs/1905.02244)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).