Hugging Face's logo

qualcomm
/
Real-ESRGAN-General-x4v3

Image-to-Image
PyTorch
LiteRT
android
Model card Files
xet
Community

Real-ESRGAN-General-x4v3: Optimized for Mobile Deployment

Upscale images and remove image noise

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

This model is an implementation of Real-ESRGAN-General-x4v3 found here.

This repository provides scripts to run Real-ESRGAN-General-x4v3 on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.super_resolution
  • Model Stats:
    • Model checkpoint: realesr-general-x4v3
    • Input resolution: 128x128
    • Number of parameters: 1.21M
    • Model size (float): 4.65 MB
    • Model size (w8a8): 1.25 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
Real-ESRGAN-General-x4v3 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 33.957 ms 3 - 31 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 33.094 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 9.495 ms 3 - 51 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 8.867 ms 0 - 49 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 6.298 ms 0 - 8 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 5.831 ms 0 - 10 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 6.477 ms 6 - 19 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 9.889 ms 3 - 31 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 41.915 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float SA7255P ADP Qualcomm® SA7255P TFLITE 33.957 ms 3 - 31 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float SA7255P ADP Qualcomm® SA7255P QNN_DLC 33.094 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 6.306 ms 0 - 15 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 5.833 ms 0 - 11 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float SA8295P ADP Qualcomm® SA8295P TFLITE 11.016 ms 3 - 39 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float SA8295P ADP Qualcomm® SA8295P QNN_DLC 9.727 ms 0 - 38 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 6.289 ms 2 - 17 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 5.828 ms 0 - 10 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float SA8775P ADP Qualcomm® SA8775P TFLITE 9.889 ms 3 - 31 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float SA8775P ADP Qualcomm® SA8775P QNN_DLC 41.915 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 4.63 ms 0 - 41 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 4.273 ms 0 - 35 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 4.774 ms 6 - 50 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 3.571 ms 0 - 34 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 3.227 ms 0 - 35 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 3.775 ms 0 - 29 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 2.91 ms 0 - 30 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 2.374 ms 0 - 36 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 2.986 ms 1 - 28 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 6.298 ms 11 - 11 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 6.423 ms 8 - 8 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 TFLITE 5.906 ms 1 - 5 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 6.239 ms 0 - 103 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 152.918 ms 23 - 29 MB CPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 5.908 ms 1 - 28 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 5.662 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 2.844 ms 0 - 36 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 2.712 ms 0 - 37 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 1.874 ms 0 - 8 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 1.745 ms 0 - 9 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 1.933 ms 0 - 12 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 2.177 ms 0 - 27 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 2.056 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 35.498 ms 1 - 3 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 113.182 ms 23 - 26 MB CPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 5.908 ms 1 - 28 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 5.662 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 1.869 ms 0 - 8 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 1.749 ms 0 - 8 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 3.328 ms 0 - 33 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 3.263 ms 0 - 34 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 1.869 ms 0 - 9 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 1.749 ms 0 - 8 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 2.177 ms 0 - 27 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 2.056 ms 0 - 26 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 1.314 ms 0 - 40 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 1.235 ms 0 - 37 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 1.336 ms 0 - 33 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 1.059 ms 0 - 39 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.919 ms 0 - 33 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 1.107 ms 0 - 33 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile TFLITE 2.593 ms 0 - 34 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 2.428 ms 0 - 32 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 140.458 ms 26 - 44 MB CPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.826 ms 0 - 30 MB NPU Real-ESRGAN-General-x4v3.tflite
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.655 ms 0 - 28 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.862 ms 0 - 34 MB NPU Real-ESRGAN-General-x4v3.onnx.zip
Real-ESRGAN-General-x4v3 w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1.963 ms 10 - 10 MB NPU Real-ESRGAN-General-x4v3.dlc
Real-ESRGAN-General-x4v3 w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.879 ms 2 - 2 MB NPU Real-ESRGAN-General-x4v3.onnx.zip

Installation

Install the package via pip:

# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install "qai-hub-models[real-esrgan-general-x4v3]"

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

Sign-in to Qualcomm® AI Hub Workbench 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.

qai-hub configure --api_token API_TOKEN

Navigate to 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.

python -m qai_hub_models.models.real_esrgan_general_x4v3.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.real_esrgan_general_x4v3.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.
python -m qai_hub_models.models.real_esrgan_general_x4v3.export

How does this work?

This export script leverages Qualcomm® AI Hub 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.

import torch

import qai_hub as hub
from qai_hub_models.models.real_esrgan_general_x4v3 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.

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.

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.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.real_esrgan_general_x4v3.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.real_esrgan_general_x4v3.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Real-ESRGAN-General-x4v3's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Real-ESRGAN-General-x4v3 can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
211
Inference Providers NEW
Image-to-Image
This model isn't deployed by any Inference Provider. 🙋 1 Ask for provider support

Paper for qualcomm/Real-ESRGAN-General-x4v3