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-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® X2 Elite	1.275 ms	7 - 7 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	2.076 ms	68 - 68 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.469 ms	0 - 280 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	2.753 ms	1 - 219 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	2.76 ms	7 - 17 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.327 ms	1 - 3 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8450	2.753 ms	1 - 219 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.775 ms	7 - 17 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	2.076 ms	68 - 68 MB	NPU
decoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	1.286 ms	0 - 192 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Mobile	1.286 ms	0 - 192 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.183 ms	2 - 179 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	1.636 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	2.26 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.397 ms	0 - 277 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	2.678 ms	2 - 218 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	2.68 ms	7 - 16 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	4.174 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.007 ms	1 - 2 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8775P	2.855 ms	5 - 107 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8650P	2.855 ms	5 - 107 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8255P	2.855 ms	5 - 107 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450	2.678 ms	2 - 218 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	2.26 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.271 ms	0 - 186 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	4.174 ms	5 - 105 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	2.635 ms	0 - 42 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.271 ms	0 - 186 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.145 ms	1 - 175 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.386 ms	0 - 281 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	2.447 ms	0 - 217 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	2.682 ms	0 - 84 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	4.201 ms	0 - 104 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.983 ms	0 - 2 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8775P	2.85 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8650P	2.85 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8255P	2.85 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8450	2.447 ms	0 - 217 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2.588 ms	0 - 83 MB	NPU
decoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.265 ms	0 - 191 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	4.201 ms	0 - 104 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8295P	2.679 ms	0 - 42 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Mobile	1.265 ms	0 - 191 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.145 ms	0 - 176 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	4.57 ms	16 - 16 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	11.222 ms	48 - 48 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	7.819 ms	1 - 200 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 1 Mobile	19.732 ms	16 - 334 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	13.977 ms	15 - 20 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.913 ms	0 - 57 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8450	19.732 ms	16 - 334 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	13.459 ms	15 - 19 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	11.222 ms	48 - 48 MB	NPU
encoder	ONNX	float	Qualcomm® Dragonwing™ Q-8750	5.27 ms	16 - 142 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Mobile	5.27 ms	16 - 142 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	4.17 ms	0 - 126 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	5.129 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	12.049 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	8.023 ms	2 - 191 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	19.505 ms	0 - 308 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	14.315 ms	2 - 13 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	37.105 ms	2 - 107 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	11.288 ms	2 - 23 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8775P	13.982 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8650P	13.982 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8255P	13.982 ms	2 - 109 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450	19.505 ms	0 - 308 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	13.807 ms	2 - 12 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	12.049 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	5.444 ms	2 - 115 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	37.105 ms	2 - 107 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	19.684 ms	2 - 234 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Mobile	5.444 ms	2 - 115 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	4.356 ms	2 - 116 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	7.708 ms	6 - 198 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	19.617 ms	7 - 312 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	13.99 ms	7 - 67 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	36.844 ms	7 - 119 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.684 ms	7 - 9 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8775P	13.488 ms	7 - 110 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8650P	13.488 ms	7 - 110 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8255P	13.488 ms	7 - 110 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8450	19.617 ms	7 - 312 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	13.343 ms	6 - 66 MB	NPU
encoder	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	5.206 ms	5 - 118 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	36.844 ms	7 - 119 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8295P	19.074 ms	7 - 237 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Mobile	5.206 ms	5 - 118 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	4.099 ms	6 - 113 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