LaMa-Dilated: Optimized for Qualcomm Devices

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

This is based on the implementation of LaMa-Dilated found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.

Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit LaMa-Dilated on Qualcomm® AI Hub.

Option 2: Export with Custom Configurations

Use the Qualcomm® AI Hub Models 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 LaMa-Dilated on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_editing

Model Stats:

• Input resolution: 512x512
• Model checkpoint: Dilated CelebAHQ
• Model size (float): 174 MB
• Number of parameters: 45.6M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
LaMa-Dilated	ONNX	float	Snapdragon® X2 Elite	36.276 ms	8 - 8 MB	NPU
LaMa-Dilated	ONNX	float	Snapdragon® X Elite	77.539 ms	90 - 90 MB	NPU
LaMa-Dilated	ONNX	float	Snapdragon® 8 Gen 3 Mobile	55.112 ms	1 - 480 MB	NPU
LaMa-Dilated	ONNX	float	Snapdragon® 8 Gen 1 Mobile	137.281 ms	11 - 540 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	123.656 ms	6 - 14 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	77.963 ms	0 - 94 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® QCS8450	137.281 ms	11 - 540 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	108.998 ms	6 - 13 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	77.539 ms	90 - 90 MB	NPU
LaMa-Dilated	ONNX	float	Qualcomm® Dragonwing™ Q-8750	43.582 ms	9 - 438 MB	NPU
LaMa-Dilated	ONNX	float	Snapdragon® 8 Elite Mobile	43.582 ms	9 - 438 MB	NPU
LaMa-Dilated	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	34.329 ms	7 - 407 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® X2 Elite	34.284 ms	4 - 4 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® X Elite	75.657 ms	4 - 4 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	52.604 ms	0 - 454 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	133.18 ms	4 - 408 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	120.715 ms	4 - 13 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	403.039 ms	1 - 364 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	75.099 ms	4 - 6 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® SA8775P	106.999 ms	1 - 364 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® SA8650P	106.999 ms	1 - 364 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® SA8255P	106.999 ms	1 - 364 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® QCS8450	133.18 ms	4 - 408 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	106.383 ms	4 - 13 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	75.657 ms	4 - 4 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	40.476 ms	1 - 368 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® SA7255P	403.039 ms	1 - 364 MB	NPU
LaMa-Dilated	QNN_DLC	float	Qualcomm® SA8295P	109.466 ms	1 - 300 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® 8 Elite Mobile	40.476 ms	1 - 368 MB	NPU
LaMa-Dilated	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	32.413 ms	1 - 391 MB	NPU
LaMa-Dilated	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	52.802 ms	2 - 538 MB	NPU
LaMa-Dilated	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	136.119 ms	3 - 464 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	121.134 ms	3 - 106 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	404.325 ms	3 - 382 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	74.741 ms	0 - 130 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® SA8775P	106.505 ms	3 - 381 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® SA8650P	106.505 ms	3 - 381 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® SA8255P	106.505 ms	3 - 381 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® QCS8450	136.119 ms	3 - 464 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	110.817 ms	1 - 103 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	40.664 ms	3 - 393 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® SA7255P	404.325 ms	3 - 382 MB	NPU
LaMa-Dilated	TFLITE	float	Qualcomm® SA8295P	110.145 ms	3 - 311 MB	NPU
LaMa-Dilated	TFLITE	float	Snapdragon® 8 Elite Mobile	40.664 ms	3 - 393 MB	NPU
LaMa-Dilated	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	32.486 ms	2 - 415 MB	NPU

License

• The license for the original implementation of LaMa-Dilated can be found here.

References

• Resolution-robust Large Mask Inpainting with Fourier Convolutions
• Source Model Implementation

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.