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 2.45, 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 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-stage1
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® X2 Elite	1.273 ms	7 - 7 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	2.072 ms	68 - 68 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.454 ms	0 - 283 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.723 ms	1 - 219 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	2.793 ms	7 - 17 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.112 ms	1 - 3 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8450	2.723 ms	1 - 219 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.783 ms	7 - 16 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	2.072 ms	68 - 68 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	1.307 ms	0 - 195 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Mobile	1.307 ms	0 - 195 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.187 ms	2 - 179 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	1.61 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	2.223 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.383 ms	0 - 274 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	2.678 ms	2 - 217 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	2.639 ms	7 - 16 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	4.17 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.016 ms	4 - 177 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8775P	2.886 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8650P	2.886 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8255P	2.886 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450	2.678 ms	2 - 217 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	2.645 ms	7 - 16 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	2.223 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.25 ms	0 - 188 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	4.17 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	2.621 ms	0 - 41 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.25 ms	0 - 188 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.152 ms	1 - 173 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.375 ms	0 - 280 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2.443 ms	0 - 218 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	2.678 ms	0 - 84 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	4.215 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.952 ms	0 - 2 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8775P	2.838 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8650P	2.838 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8255P	2.838 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8450	2.443 ms	0 - 218 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2.626 ms	0 - 83 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.234 ms	0 - 189 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	4.215 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Mobile	1.234 ms	0 - 189 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.129 ms	0 - 177 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	4.549 ms	16 - 16 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	11.422 ms	47 - 47 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	7.835 ms	0 - 200 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	19.866 ms	0 - 317 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	13.966 ms	15 - 20 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.914 ms	0 - 57 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8450	19.866 ms	0 - 317 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	13.355 ms	15 - 19 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	11.422 ms	47 - 47 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	5.282 ms	16 - 144 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Mobile	5.282 ms	16 - 144 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	4.167 ms	0 - 126 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	5.224 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	12.056 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	8.007 ms	0 - 188 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	19.486 ms	0 - 306 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	14.267 ms	2 - 13 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	37.071 ms	2 - 107 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	11.286 ms	2 - 3 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8775P	14.013 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8650P	14.013 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8255P	14.013 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450	19.486 ms	0 - 306 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	13.793 ms	2 - 12 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	12.056 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	5.448 ms	2 - 116 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	37.071 ms	2 - 107 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	19.696 ms	2 - 234 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	5.448 ms	2 - 116 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	4.356 ms	2 - 117 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	7.703 ms	0 - 189 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	19.663 ms	5 - 310 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	13.696 ms	7 - 67 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	36.837 ms	7 - 119 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.72 ms	7 - 9 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8775P	13.486 ms	7 - 111 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8650P	13.486 ms	7 - 111 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8255P	13.486 ms	7 - 111 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8450	19.663 ms	5 - 310 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	13.352 ms	7 - 66 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	5.202 ms	7 - 120 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	36.837 ms	7 - 119 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Mobile	5.202 ms	7 - 120 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	4.118 ms	7 - 111 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.

Downloads last month: -; Downloads are not tracked for this model. How to track

Space using qualcomm/TrOCR 1

Paper for qualcomm/TrOCR

TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models

Paper • 2109.10282 • Published Sep 21, 2021 • 13