TrOCR: Optimized for Qualcomm Devices

End-to-end text recognition approach with pre-trained image transformer and text transformer models for both image understanding and wordpiece-level text generation.

This is based on the implementation of TrOCR 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 2.50, ONNX Runtime 1.30.0	Download
QNN_DLC	float	Universal	QAIRT 2.50	Download
TFLITE	float	Universal	QAIRT 2.50	Download

For more device-specific assets and performance metrics, visit TrOCR 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 TrOCR on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.image_to_text

Model Stats:

Input resolution: 320x320
Model checkpoint: trocr-small-handwritten
Model size (decoder) (float): 146 MB
Model size (encoder) (float): 87.8 MB
Number of parameters (decoder): 38.3M
Number of parameters (encoder): 23.0M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	1.282 ms	2 - 178 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.237 ms	0 - 191 MB	NPU
decoder	ONNX	float	Snapdragon® X2 Elite	1.251 ms	7 - 7 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	2.029 ms	68 - 68 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.479 ms	0 - 283 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.351 ms	1 - 224 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	2.8 ms	7 - 18 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.227 ms	0 - 3 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8450	2.351 ms	1 - 224 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.73 ms	7 - 17 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	2.029 ms	68 - 68 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	1.237 ms	0 - 191 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	1.268 ms	1 - 176 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.27 ms	0 - 189 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	1.595 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	2.138 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.466 ms	0 - 280 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	2.276 ms	2 - 222 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	2.677 ms	7 - 17 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.012 ms	2 - 4 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8775P	2.819 ms	7 - 112 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8650P	2.819 ms	7 - 112 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8255P	2.819 ms	7 - 112 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450	2.276 ms	2 - 222 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	2.67 ms	9 - 18 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	2.138 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.27 ms	0 - 189 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	4.14 ms	7 - 110 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	2.591 ms	0 - 45 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	1.271 ms	0 - 175 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.22 ms	0 - 191 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.437 ms	0 - 282 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2.285 ms	0 - 223 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	2.669 ms	0 - 84 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.968 ms	0 - 2 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8450	2.285 ms	0 - 223 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2.596 ms	0 - 83 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.22 ms	0 - 191 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	4.148 ms	0 - 105 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8295P	2.5 ms	0 - 46 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.571 ms	2 - 129 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	5.604 ms	16 - 143 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	4.93 ms	16 - 16 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	12.126 ms	48 - 48 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	8.089 ms	0 - 206 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	19.029 ms	1 - 319 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	14.323 ms	15 - 20 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	11.516 ms	15 - 18 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8450	19.029 ms	1 - 319 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	14.312 ms	15 - 20 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	12.126 ms	48 - 48 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	5.604 ms	16 - 143 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.256 ms	2 - 116 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	5.267 ms	2 - 113 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	5.116 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	11.621 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	7.686 ms	2 - 194 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	17.937 ms	0 - 303 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	13.719 ms	2 - 13 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.91 ms	2 - 171 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8775P	13.448 ms	2 - 110 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8650P	13.448 ms	2 - 110 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8255P	13.448 ms	2 - 110 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450	17.937 ms	0 - 303 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	13.638 ms	2 - 12 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	11.621 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	5.267 ms	2 - 113 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	36.062 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	18.993 ms	2 - 222 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	4.05 ms	0 - 108 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	4.993 ms	6 - 111 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	7.382 ms	5 - 196 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	17.205 ms	7 - 307 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	13.228 ms	7 - 67 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.294 ms	7 - 9 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8450	17.205 ms	7 - 307 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	12.847 ms	6 - 66 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	4.993 ms	6 - 111 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	35.353 ms	7 - 111 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8295P	18.406 ms	7 - 222 MB	NPU

License

The license for the original implementation of TrOCR 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.

qualcomm/TrOCR