Hugging Face's logo

qualcomm
/
AOT-GAN

Image-to-Image
PyTorch
LiteRT
android
Model card Files
xet
Community
1

AOT-GAN: Optimized for Mobile Deployment

High resolution image in-painting on-device

AOT-GAN is a machine learning model that allows to erase and in-paint part of given input image.

This model is an implementation of AOT-GAN found here.

This repository provides scripts to run AOT-GAN on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.image_editing
  • Model Stats:
    • Model checkpoint: CelebAHQ
    • Input resolution: 512x512
    • Number of parameters: 15.2M
    • Model size (float): 58.0 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
AOT-GAN float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 635.313 ms 3 - 141 MB NPU AOT-GAN.tflite
AOT-GAN float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 548.886 ms 0 - 257 MB NPU AOT-GAN.dlc
AOT-GAN float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 203.851 ms 3 - 157 MB NPU AOT-GAN.tflite
AOT-GAN float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 216.398 ms 4 - 219 MB NPU AOT-GAN.dlc
AOT-GAN float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 161.319 ms 3 - 30 MB NPU AOT-GAN.tflite
AOT-GAN float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 121.76 ms 4 - 84 MB NPU AOT-GAN.dlc
AOT-GAN float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 118.772 ms 0 - 146 MB NPU AOT-GAN.onnx.zip
AOT-GAN float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 211.172 ms 3 - 142 MB NPU AOT-GAN.tflite
AOT-GAN float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 771.087 ms 1 - 261 MB NPU AOT-GAN.dlc
AOT-GAN float SA7255P ADP Qualcomm® SA7255P TFLITE 635.313 ms 3 - 141 MB NPU AOT-GAN.tflite
AOT-GAN float SA7255P ADP Qualcomm® SA7255P QNN_DLC 548.886 ms 0 - 257 MB NPU AOT-GAN.dlc
AOT-GAN float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 161.458 ms 0 - 27 MB NPU AOT-GAN.tflite
AOT-GAN float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 120.895 ms 4 - 69 MB NPU AOT-GAN.dlc
AOT-GAN float SA8295P ADP Qualcomm® SA8295P TFLITE 231.233 ms 3 - 132 MB NPU AOT-GAN.tflite
AOT-GAN float SA8295P ADP Qualcomm® SA8295P QNN_DLC 182.791 ms 1 - 226 MB NPU AOT-GAN.dlc
AOT-GAN float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 164.246 ms 1 - 29 MB NPU AOT-GAN.tflite
AOT-GAN float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 121.997 ms 3 - 74 MB NPU AOT-GAN.dlc
AOT-GAN float SA8775P ADP Qualcomm® SA8775P TFLITE 211.172 ms 3 - 142 MB NPU AOT-GAN.tflite
AOT-GAN float SA8775P ADP Qualcomm® SA8775P QNN_DLC 771.087 ms 1 - 261 MB NPU AOT-GAN.dlc
AOT-GAN float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 119.536 ms 0 - 161 MB NPU AOT-GAN.tflite
AOT-GAN float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 89.16 ms 1 - 260 MB NPU AOT-GAN.dlc
AOT-GAN float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 86.116 ms 8 - 259 MB NPU AOT-GAN.onnx.zip
AOT-GAN float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 101.077 ms 2 - 143 MB NPU AOT-GAN.tflite
AOT-GAN float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 67.565 ms 1 - 279 MB NPU AOT-GAN.dlc
AOT-GAN float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 67.545 ms 8 - 294 MB NPU AOT-GAN.onnx.zip
AOT-GAN float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 82.02 ms 2 - 141 MB NPU AOT-GAN.tflite
AOT-GAN float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 47.483 ms 4 - 231 MB NPU AOT-GAN.dlc
AOT-GAN float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 46.537 ms 4 - 235 MB NPU AOT-GAN.onnx.zip
AOT-GAN float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 125.987 ms 138 - 138 MB NPU AOT-GAN.dlc
AOT-GAN float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 121.825 ms 32 - 32 MB NPU AOT-GAN.onnx.zip

Installation

Install the package via pip:

pip install qai-hub-models

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

Sign-in to Qualcomm® AI Hub 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.aotgan.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.aotgan.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.aotgan.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.aotgan 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. 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.aotgan.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.aotgan.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 AOT-GAN's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of AOT-GAN can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month
577
Inference Providers NEW
Image-to-Image
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support