Hugging Face's logo

qualcomm
/
XLSR

Image-to-Image
PyTorch
android
Model card Files
xet
 Community
1
XLSR
File size: 11,810 Bytes
96531c5
 
 
 
 
5130323
96531c5
 
 
fbc84ae
96531c5
ff84694
85bce41
96531c5
 
ff84694
d9dbac7
ff84694
 
 
 
 
 
 
 
 
 
 
 
d9dbac7
 
 
 
 
 
ff84694
 
 
 
 
 
d9dbac7
ff84694
 
 
 
 
 
d9dbac7
ff84694
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d9dbac7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a6b1c63
d9dbac7
 
 
 
 
 
 
 
 
a6b1c63
d9dbac7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8f9afb6
96531c5
8a76801
bb2caa6
8f9afb6
96531c5
 
 
 
 
f8cd1cc
96531c5
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
---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-to-image

---

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

# XLSR: Optimized for Qualcomm Devices

XLSR is designed for lightweight real-time upscaling of images.

This is based on the implementation of XLSR found [here](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/xlsr).
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/xlsr) 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/xlsr/releases/v0.59.0/xlsr-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/xlsr/releases/v0.59.0/xlsr-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/xlsr/releases/v0.59.0/xlsr-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/xlsr/releases/v0.59.0/xlsr-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/xlsr/releases/v0.59.0/xlsr-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/xlsr/releases/v0.59.0/xlsr-tflite-w8a8.zip)

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


### 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/xlsr) 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 [XLSR on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/xlsr) for usage instructions.

## Model Details

**Model Type:** Model_use_case.super_resolution

**Model Stats:**
- Model checkpoint: xlsr_3x_checkpoint
- Input resolution: 128x128
- Number of parameters: 28.0K
- Model size (float): 115 KB
- Model size (w8a8): 45.6 KB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| XLSR | ONNX | float | Snapdragon® X2 Elite | 0.363 ms | 8 - 8 MB | NPU
| XLSR | ONNX | float | Snapdragon® X Elite | 0.864 ms | 9 - 9 MB | NPU
| XLSR | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.576 ms | 0 - 29 MB | NPU
| XLSR | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 1.237 ms | 2 - 35 MB | NPU
| XLSR | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.873 ms | 0 - 2 MB | NPU
| XLSR | ONNX | float | Qualcomm® QCS8450 | 1.237 ms | 2 - 35 MB | NPU
| XLSR | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 1.156 ms | 7 - 10 MB | NPU
| XLSR | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.864 ms | 9 - 9 MB | NPU
| XLSR | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 0.468 ms | 0 - 19 MB | NPU
| XLSR | ONNX | float | Snapdragon® 8 Elite Mobile | 0.468 ms | 0 - 19 MB | NPU
| XLSR | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.403 ms | 0 - 19 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® X2 Elite | 0.26 ms | 3 - 3 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® X Elite | 0.642 ms | 3 - 3 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.423 ms | 0 - 33 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.728 ms | 0 - 38 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 3.293 ms | 2 - 4 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.644 ms | 0 - 2 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® QCS8450 | 0.728 ms | 0 - 38 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.779 ms | 0 - 3 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.642 ms | 3 - 3 MB | NPU
| XLSR | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.308 ms | 0 - 28 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 0.308 ms | 0 - 28 MB | NPU
| XLSR | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.278 ms | 0 - 23 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® X2 Elite | 0.408 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® X Elite | 0.855 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.46 ms | 0 - 29 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 1.051 ms | 0 - 35 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 2.188 ms | 0 - 21 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.776 ms | 0 - 2 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® SA8775P | 1.138 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® SA8650P | 1.138 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® SA8255P | 1.138 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® QCS8450 | 1.051 ms | 0 - 35 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 1.067 ms | 0 - 5 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 0.855 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 0.363 ms | 0 - 26 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® SA7255P | 2.188 ms | 0 - 21 MB | NPU
| XLSR | QNN_DLC | float | Qualcomm® SA8295P | 1.408 ms | 0 - 18 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 0.363 ms | 0 - 26 MB | NPU
| XLSR | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.339 ms | 0 - 24 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.256 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.481 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.251 ms | 0 - 29 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.467 ms | 0 - 30 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 1.419 ms | 0 - 2 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 0.922 ms | 0 - 21 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.402 ms | 0 - 27 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.581 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® SA8650P | 0.581 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® SA8255P | 0.581 ms | 0 - 22 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 0.467 ms | 0 - 30 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.512 ms | 0 - 2 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 0.481 ms | 0 - 0 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1.594 ms | 0 - 19 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.412 ms | 0 - 20 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.209 ms | 0 - 26 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® SA7255P | 0.922 ms | 0 - 21 MB | NPU
| XLSR | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.779 ms | 0 - 18 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 0.209 ms | 0 - 26 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.186 ms | 0 - 23 MB | NPU
| XLSR | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.412 ms | 0 - 20 MB | NPU
| XLSR | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.395 ms | 0 - 31 MB | NPU
| XLSR | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 2.517 ms | 0 - 36 MB | NPU
| XLSR | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 4.683 ms | 3 - 24 MB | NPU
| XLSR | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 2.307 ms | 0 - 1 MB | NPU
| XLSR | TFLITE | float | Qualcomm® SA8775P | 2.642 ms | 0 - 23 MB | NPU
| XLSR | TFLITE | float | Qualcomm® SA8650P | 2.642 ms | 0 - 23 MB | NPU
| XLSR | TFLITE | float | Qualcomm® SA8255P | 2.642 ms | 0 - 23 MB | NPU
| XLSR | TFLITE | float | Qualcomm® QCS8450 | 2.517 ms | 0 - 36 MB | NPU
| XLSR | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 2.501 ms | 1 - 6 MB | NPU
| XLSR | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 0.963 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | float | Qualcomm® SA7255P | 4.683 ms | 3 - 24 MB | NPU
| XLSR | TFLITE | float | Qualcomm® SA8295P | 3.063 ms | 0 - 18 MB | NPU
| XLSR | TFLITE | float | Snapdragon® 8 Elite Mobile | 0.963 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.796 ms | 0 - 24 MB | NPU
| XLSR | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.253 ms | 0 - 29 MB | NPU
| XLSR | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 0.472 ms | 0 - 36 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 1.43 ms | 0 - 3 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8275 | 0.982 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 0.403 ms | 0 - 2 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® SA8775P | 0.607 ms | 0 - 22 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® SA8650P | 0.607 ms | 0 - 22 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® SA8255P | 0.607 ms | 0 - 22 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® QCS8450 | 0.472 ms | 0 - 36 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 0.545 ms | 0 - 3 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1.618 ms | 0 - 19 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 0.416 ms | 0 - 20 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 0.23 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® SA7255P | 0.982 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | w8a8 | Qualcomm® SA8295P | 0.797 ms | 0 - 18 MB | NPU
| XLSR | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 0.23 ms | 0 - 21 MB | NPU
| XLSR | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.191 ms | 0 - 22 MB | NPU
| XLSR | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.416 ms | 0 - 20 MB | NPU

## License
* The license for the original implementation of XLSR can be found
  [here](https://github.com/quic/aimet-model-zoo/blob/develop/LICENSE.md).

## References
* [Extremely Lightweight Quantization Robust Real-Time Single-Image Super Resolution for Mobile Devices](https://arxiv.org/abs/2105.10288)
* [Source Model Implementation](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/xlsr)

## 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).