LaMa-Dilated: Optimized for Mobile Deployment
High resolution image in-painting on-device
LaMa-Dilated is a machine learning model that allows to erase and in-paint part of given input image.
This model is an implementation of LaMa-Dilated found here.
This repository provides scripts to run LaMa-Dilated on Qualcomm® devices. More details on model performance across various devices, can be found here.
Model Details
- Model Type: Image editing
- Model Stats:
- Model checkpoint: Dilated CelebAHQ
- Input resolution: 512x512
- Number of parameters: 45.6M
- Model size: 174 MB
Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |
---|---|---|---|---|---|---|---|---|
LaMa-Dilated | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 75.011 ms | 3 - 322 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 70.796 ms | 3 - 38 MB | FP16 | NPU | LaMa-Dilated.so |
LaMa-Dilated | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 73.503 ms | 0 - 1247 MB | FP16 | NPU | LaMa-Dilated.onnx |
LaMa-Dilated | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 52.544 ms | 3 - 92 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 48.034 ms | 4 - 89 MB | FP16 | NPU | LaMa-Dilated.so |
LaMa-Dilated | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 50.628 ms | 0 - 267 MB | FP16 | NPU | LaMa-Dilated.onnx |
LaMa-Dilated | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 49.125 ms | 2 - 92 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 45.679 ms | 4 - 91 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 47.643 ms | 0 - 164 MB | FP16 | NPU | LaMa-Dilated.onnx |
LaMa-Dilated | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 75.063 ms | 3 - 321 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 66.641 ms | 4 - 5 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | SA7255P ADP | SA7255P | TFLITE | 3020.421 ms | 0 - 89 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | SA7255P ADP | SA7255P | QNN | 3006.571 ms | 3 - 13 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 74.93 ms | 3 - 334 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | SA8255 (Proxy) | SA8255P Proxy | QNN | 66.578 ms | 4 - 6 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | SA8295P ADP | SA8295P | TFLITE | 115.892 ms | 3 - 45 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | SA8295P ADP | SA8295P | QNN | 103.929 ms | 1 - 7 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 75.69 ms | 3 - 313 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | SA8650 (Proxy) | SA8650P Proxy | QNN | 66.8 ms | 4 - 5 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | SA8775P ADP | SA8775P | TFLITE | 129.854 ms | 3 - 91 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | SA8775P ADP | SA8775P | QNN | 119.068 ms | 2 - 8 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 107.83 ms | 3 - 48 MB | FP16 | NPU | LaMa-Dilated.tflite |
LaMa-Dilated | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 99.617 ms | 4 - 47 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 69.489 ms | 4 - 4 MB | FP16 | NPU | Use Export Script |
LaMa-Dilated | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 78.52 ms | 90 - 90 MB | FP16 | NPU | LaMa-Dilated.onnx |
Installation
This model can be installed as a Python package via pip.
pip install "qai-hub-models[lama_dilated]"
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.lama_dilated.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.lama_dilated.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.lama_dilated.export
Profiling Results
------------------------------------------------------------
LaMa-Dilated
Device : Samsung Galaxy S23 (13)
Runtime : TFLITE
Estimated inference time (ms) : 75.0
Estimated peak memory usage (MB): [3, 322]
Total # Ops : 343
Compute Unit(s) : NPU (343 ops)
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.lama_dilated import Model
# Load the model
torch_model = Model.from_pretrained()
# Device
device = hub.Device("Samsung Galaxy S23")
# 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.lama_dilated.demo --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.lama_dilated.demo -- --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 LaMa-Dilated's performance across various devices here. Explore all available models on Qualcomm® AI Hub
License
- The license for the original implementation of LaMa-Dilated can be found here.
- The license for the compiled assets for on-device deployment can be found here
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.