Hugging Face's logo

qualcomm
/
RegNet

Image Classification
PyTorch
backbone
android
Model card Files
xet
 Community
RegNet
File size: 12,117 Bytes
a97ffba
 
0e88ee3
a97ffba
 
 
abcd9fe
a97ffba
 
 
8c697f6
a97ffba
8c067c7
ea567a9
a97ffba
 
8c067c7
f12ed77
8c067c7
 
 
 
 
 
 
 
 
 
 
 
f12ed77
 
 
 
 
 
8c067c7
 
 
 
 
 
f12ed77
8c067c7
 
 
 
 
 
f12ed77
8c067c7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f12ed77
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8845495
f12ed77
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8845495
f12ed77
 
 
 
 
 
c20270d
a97ffba
4489a96
 
c20270d
a97ffba
 
 
 
 
82e0e91
a97ffba
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
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
---
library_name: pytorch
license: other
tags:
- backbone
- android
pipeline_tag: image-classification

---

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

# RegNet: Optimized for Qualcomm Devices

RegNet 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 RegNet found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/regnet.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/v0.59.0/src/qai_hub_models/models/regnet) 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 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/regnet/releases/v0.59.0/regnet-tflite-w8a8.zip)

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


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/regnet) 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 [RegNet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/regnet) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 15.3M
- Model size (float): 58.3 MB
- Model size (w8a8): 15.4 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| RegNet | ONNX | float | Snapdragon® X2 Elite | 0.911 ms | 2 - 2 MB | NPU
| RegNet | ONNX | float | Snapdragon® X Elite | 1.772 ms | 39 - 39 MB | NPU
| RegNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.229 ms | 0 - 127 MB | NPU
| RegNet | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 2.731 ms | 0 - 116 MB | NPU
| RegNet | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 1.746 ms | 0 - 42 MB | NPU
| RegNet | ONNX | float | Qualcomm® QCS8450 | 2.731 ms | 0 - 116 MB | NPU
| RegNet | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 2.809 ms | 1 - 4 MB | NPU
| RegNet | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 1.772 ms | 39 - 39 MB | NPU
| RegNet | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.995 ms | 0 - 75 MB | NPU
| RegNet | ONNX | float | Snapdragon® 8 Elite Mobile | 0.995 ms | 0 - 75 MB | NPU
| RegNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.837 ms | 0 - 77 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® X2 Elite | 0.472 ms | 1 - 1 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® X Elite | 0.928 ms | 20 - 20 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.642 ms | 0 - 135 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.292 ms | 0 - 134 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 4.221 ms | 0 - 3 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.901 ms | 0 - 170 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® QCS8450 | 1.292 ms | 0 - 134 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.159 ms | 0 - 3 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.928 ms | 20 - 20 MB | NPU
| RegNet | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.549 ms | 0 - 90 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.549 ms | 0 - 90 MB | NPU
| RegNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.507 ms | 0 - 90 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® X2 Elite | 1.204 ms | 1 - 1 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® X Elite | 2.288 ms | 1 - 1 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.414 ms | 0 - 127 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 3.501 ms | 0 - 116 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 9.87 ms | 1 - 71 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.055 ms | 1 - 2 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® SA8775P | 3.275 ms | 0 - 73 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® SA8650P | 3.275 ms | 0 - 73 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® SA8255P | 3.275 ms | 0 - 73 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® QCS8450 | 3.501 ms | 0 - 116 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 3.044 ms | 1 - 3 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 2.288 ms | 1 - 1 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 1.114 ms | 0 - 76 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® SA7255P | 9.87 ms | 1 - 71 MB | NPU
| RegNet | QNN_DLC | float | Qualcomm® SA8295P | 3.472 ms | 0 - 61 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.114 ms | 0 - 76 MB | NPU
| RegNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.901 ms | 1 - 76 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.568 ms | 0 - 0 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 1.072 ms | 0 - 0 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.632 ms | 0 - 111 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.307 ms | 0 - 111 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.962 ms | 0 - 2 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 2.296 ms | 0 - 80 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.896 ms | 0 - 94 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® SA8775P | 1.291 ms | 0 - 82 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® SA8650P | 1.291 ms | 0 - 82 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® SA8255P | 1.291 ms | 0 - 82 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 1.307 ms | 0 - 111 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 1.114 ms | 0 - 2 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 1.072 ms | 0 - 0 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 7.137 ms | 0 - 199 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 1.178 ms | 0 - 80 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.486 ms | 0 - 78 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® SA7255P | 2.296 ms | 0 - 80 MB | NPU
| RegNet | QNN_DLC | w8a8 | Qualcomm® SA8295P | 1.609 ms | 0 - 78 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.486 ms | 0 - 78 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.443 ms | 0 - 83 MB | NPU
| RegNet | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 1.178 ms | 0 - 80 MB | NPU
| RegNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.392 ms | 0 - 152 MB | NPU
| RegNet | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 3.481 ms | 0 - 137 MB | NPU
| RegNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 9.873 ms | 0 - 89 MB | NPU
| RegNet | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.032 ms | 0 - 2 MB | NPU
| RegNet | TFLITE | float | Qualcomm® SA8775P | 3.285 ms | 0 - 91 MB | NPU
| RegNet | TFLITE | float | Qualcomm® SA8650P | 3.285 ms | 0 - 91 MB | NPU
| RegNet | TFLITE | float | Qualcomm® SA8255P | 3.285 ms | 0 - 91 MB | NPU
| RegNet | TFLITE | float | Qualcomm® QCS8450 | 3.481 ms | 0 - 137 MB | NPU
| RegNet | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 3.043 ms | 0 - 42 MB | NPU
| RegNet | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 1.107 ms | 0 - 93 MB | NPU
| RegNet | TFLITE | float | Qualcomm® SA7255P | 9.873 ms | 0 - 89 MB | NPU
| RegNet | TFLITE | float | Qualcomm® SA8295P | 3.467 ms | 0 - 72 MB | NPU
| RegNet | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.107 ms | 0 - 93 MB | NPU
| RegNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.899 ms | 0 - 96 MB | NPU
| RegNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.524 ms | 0 - 118 MB | NPU
| RegNet | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 1.118 ms | 0 - 116 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.846 ms | 0 - 22 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 1.982 ms | 0 - 75 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.719 ms | 0 - 2 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® SA8775P | 1.109 ms | 0 - 79 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® SA8650P | 1.109 ms | 0 - 79 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® SA8255P | 1.109 ms | 0 - 79 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® QCS8450 | 1.118 ms | 0 - 116 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.936 ms | 0 - 22 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 6.644 ms | 0 - 192 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.988 ms | 0 - 73 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.435 ms | 0 - 81 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® SA7255P | 1.982 ms | 0 - 75 MB | NPU
| RegNet | TFLITE | w8a8 | Qualcomm® SA8295P | 1.42 ms | 0 - 72 MB | NPU
| RegNet | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.435 ms | 0 - 81 MB | NPU
| RegNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.393 ms | 0 - 76 MB | NPU
| RegNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.988 ms | 0 - 73 MB | NPU

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

## References
* [Designing Network Design Spaces](https://arxiv.org/abs/2003.13678)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/regnet.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).