README.md

---
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/esrgan/web-assets/model_demo.png)

# ESRGAN: Optimized for Mobile Deployment
## Upscale images and remove image noise


ESRGAN is a machine learning model that upscales an image with minimal loss in quality.

This model is an implementation of ESRGAN found [here](https://github.com/xinntao/ESRGAN/).


This repository provides scripts to run ESRGAN on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/esrgan).



### Model Details

- **Model Type:** Model_use_case.super_resolution
- **Model Stats:**
  - Model checkpoint: ESRGAN_x4
  - Input resolution: 128x128
  - Number of parameters: 16.7M
  - Model size (float): 63.9 MB

| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|---|---|---|---|---|---|---|---|---|
| ESRGAN | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 448.729 ms | 3 - 319 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 448.654 ms | 0 - 278 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 130.63 ms | 3 - 730 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 118.728 ms | 1 - 694 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 65.593 ms | 2 - 5 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 68.161 ms | 0 - 3 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 67.464 ms | 0 - 43 MB | NPU | [ESRGAN.onnx.zip](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.onnx.zip) |
| ESRGAN | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 105.399 ms | 3 - 318 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 496.068 ms | 0 - 277 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 448.729 ms | 3 - 319 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 448.654 ms | 0 - 278 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 111.285 ms | 3 - 331 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 111.274 ms | 0 - 297 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 105.399 ms | 3 - 318 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 496.068 ms | 0 - 277 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 50.392 ms | 3 - 728 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 48.83 ms | 0 - 685 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 50.253 ms | 8 - 698 MB | NPU | [ESRGAN.onnx.zip](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.onnx.zip) |
| ESRGAN | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 38.193 ms | 3 - 303 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 38.092 ms | 0 - 268 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 39.247 ms | 7 - 258 MB | NPU | [ESRGAN.onnx.zip](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.onnx.zip) |
| ESRGAN | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 26.103 ms | 3 - 310 MB | NPU | [ESRGAN.tflite](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.tflite) |
| ESRGAN | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 25.976 ms | 0 - 276 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 29.96 ms | 7 - 266 MB | NPU | [ESRGAN.onnx.zip](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.onnx.zip) |
| ESRGAN | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 64.764 ms | 0 - 0 MB | NPU | [ESRGAN.dlc](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.dlc) |
| ESRGAN | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 65.553 ms | 37 - 37 MB | NPU | [ESRGAN.onnx.zip](https://huggingface.co/qualcomm/ESRGAN/blob/main/ESRGAN.onnx.zip) |




## Installation


Install the package via pip:
```bash
pip install qai-hub-models
```


## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device

Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.

With this API token, you can configure your client to run models on the cloud
hosted devices.
```bash
qai-hub configure --api_token API_TOKEN
```
Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.



## Demo off target

The package contains a simple end-to-end demo that downloads pre-trained
weights and runs this model on a sample input.

```bash
python -m qai_hub_models.models.esrgan.demo
```

The above demo runs a reference implementation of pre-processing, model
inference, and post processing.

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.esrgan.demo
```


### Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
device. This script does the following:
* Performance check on-device on a cloud-hosted device
* Downloads compiled assets that can be deployed on-device for Android.
* Accuracy check between PyTorch and on-device outputs.

```bash
python -m qai_hub_models.models.esrgan.export
```



## How does this work?

This [export script](https://aihub.qualcomm.com/models/esrgan/qai_hub_models/models/ESRGAN/export.py)
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
on-device. Lets go through each step below in detail:

Step 1: **Compile model for on-device deployment**

To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the `jit.trace` and then call the `submit_compile_job` API.

```python
import torch

import qai_hub as hub
from qai_hub_models.models.esrgan import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S25")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

```


Step 2: **Performance profiling on cloud-hosted device**

After compiling models from step 1. Models can be profiled model on-device using the
`target_model`. Note that this scripts runs the model on a device automatically
provisioned in the cloud.  Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
```python
profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)
        
```

Step 3: **Verify on-device accuracy**

To verify the accuracy of the model on-device, you can run on-device inference
on sample input data on the same cloud hosted device.
```python
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

```
With the output of the model, you can compute like PSNR, relative errors or
spot check the output with expected output.

**Note**: This on-device profiling and inference requires access to Qualcomm®
AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).



## Run demo on a cloud-hosted device

You can also run the demo on-device.

```bash
python -m qai_hub_models.models.esrgan.demo --eval-mode on-device
```

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.esrgan.demo -- --eval-mode on-device
```


## Deploying compiled model to Android


The models can be deployed using multiple runtimes:
- TensorFlow Lite (`.tflite` export): [This
  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
  guide to deploy the .tflite model in an Android application.


- QNN (`.so` export ): This [sample
  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
provides instructions on how to use the `.so` shared library  in an Android application.


## View on Qualcomm® AI Hub
Get more details on ESRGAN's performance across various devices [here](https://aihub.qualcomm.com/models/esrgan).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


## License
* The license for the original implementation of ESRGAN can be found
  [here](https://github.com/xinntao/ESRGAN/blob/master/LICENSE).



## References
* [ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks](https://arxiv.org/abs/1809.00219)
* [Source Model Implementation](https://github.com/xinntao/ESRGAN/)



## 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).