| --- |
| library_name: pytorch |
| license: other |
| tags: |
| - android |
| pipeline_tag: image-to-image |
|
|
| --- |
| |
|  |
|
|
| # QuickSRNetSmall: Optimized for Qualcomm Devices |
|
|
| QuickSRNet Small is designed for upscaling images on mobile platforms to sharpen in real-time. |
|
|
| This is based on the implementation of QuickSRNetSmall found [here](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/quicksrnet). |
| 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/quicksrnetsmall) 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/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-onnx-float.zip) |
| | ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-onnx-w8a8.zip) |
| | QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-qnn_dlc-float.zip) |
| | QNN_DLC | w8a8 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-qnn_dlc-w8a8.zip) |
| | TFLITE | float | Universal | QAIRT 2.43, TFLite 2.19.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-tflite-float.zip) |
| | TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.19.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/quicksrnetsmall/releases/v0.50.2/quicksrnetsmall-tflite-w8a8.zip) |
|
|
| For more device-specific assets and performance metrics, visit **[QuickSRNetSmall on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/quicksrnetsmall)**. |
|
|
|
|
| ### 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/quicksrnetsmall) 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 [QuickSRNetSmall on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/quicksrnetsmall) for usage instructions. |
|
|
| ## Model Details |
|
|
| **Model Type:** Model_use_case.super_resolution |
| |
| **Model Stats:** |
| - Model checkpoint: quicksrnet_small_3x_checkpoint |
| - Input resolution: 128x128 |
| - Number of parameters: 33.3K |
| - Model size (float): 133 KB |
| - Model size (w8a8): 41.7 KB |
|
|
| ## Performance Summary |
| | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
| |---|---|---|---|---|---|--- |
| | QuickSRNetSmall | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.397 ms | 0 - 21 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Snapdragon® X2 Elite | 0.374 ms | 6 - 6 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Snapdragon® X Elite | 1.034 ms | 9 - 9 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.603 ms | 0 - 26 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Qualcomm® QCS8550 (Proxy) | 0.901 ms | 0 - 14 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Qualcomm® QCS9075 | 1.175 ms | 7 - 10 MB | NPU |
| | QuickSRNetSmall | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.465 ms | 0 - 17 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.216 ms | 0 - 21 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Snapdragon® X2 Elite | 0.216 ms | 3 - 3 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Snapdragon® X Elite | 0.602 ms | 3 - 3 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.338 ms | 0 - 27 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.524 ms | 0 - 2 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Qualcomm® QCS9075 | 0.669 ms | 0 - 3 MB | NPU |
| | QuickSRNetSmall | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.248 ms | 0 - 22 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.302 ms | 0 - 23 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Snapdragon® X2 Elite | 0.436 ms | 0 - 0 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Snapdragon® X Elite | 0.859 ms | 0 - 0 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 0.447 ms | 0 - 28 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 1.871 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 0.729 ms | 0 - 6 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® SA8775P | 1.087 ms | 0 - 21 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® QCS9075 | 1.105 ms | 0 - 5 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 1.025 ms | 0 - 29 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® SA7255P | 1.871 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Qualcomm® SA8295P | 1.377 ms | 0 - 17 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.344 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.148 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 0.233 ms | 0 - 0 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® X Elite | 0.425 ms | 0 - 0 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.216 ms | 0 - 24 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 1.284 ms | 2 - 4 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 0.79 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.343 ms | 0 - 1 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® SA8775P | 0.525 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 0.501 ms | 0 - 2 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 1.205 ms | 0 - 16 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.457 ms | 0 - 26 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® SA7255P | 0.79 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Qualcomm® SA8295P | 0.705 ms | 0 - 15 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.178 ms | 0 - 21 MB | NPU |
| | QuickSRNetSmall | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.351 ms | 0 - 17 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.381 ms | 0 - 23 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 0.621 ms | 0 - 28 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 2.375 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 1.012 ms | 0 - 1 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® SA8775P | 1.442 ms | 0 - 21 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® QCS9075 | 1.267 ms | 3 - 8 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 1.307 ms | 0 - 29 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® SA7255P | 2.375 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Qualcomm® SA8295P | 1.658 ms | 0 - 17 MB | NPU |
| | QuickSRNetSmall | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.443 ms | 0 - 20 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 0.161 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 0.267 ms | 0 - 25 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCS6490 | 1.144 ms | 0 - 2 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 0.891 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 0.4 ms | 0 - 1 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® SA8775P | 0.592 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCS9075 | 0.523 ms | 0 - 3 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCM6690 | 1.486 ms | 0 - 16 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 0.469 ms | 0 - 26 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® SA7255P | 0.891 ms | 0 - 19 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Qualcomm® SA8295P | 0.784 ms | 0 - 15 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 0.199 ms | 0 - 17 MB | NPU |
| | QuickSRNetSmall | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 0.408 ms | 0 - 16 MB | NPU |
| |
| ## License |
| * The license for the original implementation of QuickSRNetSmall can be found |
| [here](https://github.com/quic/aimet-model-zoo/blob/develop/LICENSE.md). |
| |
| ## References |
| * [QuickSRNet: Plain Single-Image Super-Resolution Architecture for Faster Inference on Mobile Platforms](https://arxiv.org/abs/2303.04336) |
| * [Source Model Implementation](https://github.com/quic/aimet-model-zoo/tree/develop/aimet_zoo_torch/quicksrnet) |
| |
| ## 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). |
| |
