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 2.42, ONNX Runtime 1.24.3	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:

Model checkpoint: trocr-small-stage1
Input resolution: 320x320
Number of parameters (decoder): 38.3M
Model size (decoder) (float): 146 MB
Number of parameters (encoder): 23.0M
Model size (encoder) (float): 87.8 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
decoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.155 ms	0 - 225 MB	NPU
decoder	ONNX	float	Snapdragon® X2 Elite	1.176 ms	68 - 68 MB	NPU
decoder	ONNX	float	Snapdragon® X Elite	2.299 ms	67 - 67 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.457 ms	0 - 249 MB	NPU
decoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.127 ms	0 - 76 MB	NPU
decoder	ONNX	float	Qualcomm® QCS9075	2.702 ms	7 - 16 MB	NPU
decoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.242 ms	0 - 252 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.142 ms	1 - 174 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X2 Elite	1.53 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® X Elite	2.17 ms	7 - 7 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.421 ms	0 - 281 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	4.184 ms	1 - 102 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	2.047 ms	1 - 3 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8775P	2.895 ms	0 - 102 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS9075	2.57 ms	7 - 15 MB	NPU
decoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	2.635 ms	2 - 216 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA7255P	4.184 ms	1 - 102 MB	NPU
decoder	QNN_DLC	float	Qualcomm® SA8295P	2.704 ms	7 - 49 MB	NPU
decoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.279 ms	0 - 188 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.118 ms	0 - 175 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.632 ms	0 - 281 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8275 (Proxy)	4.232 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1.986 ms	0 - 2 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8775P	2.864 ms	0 - 102 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS9075	2.537 ms	0 - 83 MB	NPU
decoder	TFLITE	float	Qualcomm® QCS8450 (Proxy)	2.444 ms	0 - 215 MB	NPU
decoder	TFLITE	float	Qualcomm® SA7255P	4.232 ms	0 - 103 MB	NPU
decoder	TFLITE	float	Qualcomm® SA8295P	2.664 ms	0 - 43 MB	NPU
decoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.229 ms	0 - 192 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	7.262 ms	16 - 326 MB	NPU
encoder	ONNX	float	Snapdragon® X2 Elite	7.526 ms	48 - 48 MB	NPU
encoder	ONNX	float	Snapdragon® X Elite	18.696 ms	48 - 48 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Gen 3 Mobile	12.569 ms	16 - 404 MB	NPU
encoder	ONNX	float	Qualcomm® QCS8550 (Proxy)	18.211 ms	0 - 61 MB	NPU
encoder	ONNX	float	Qualcomm® QCS9075	22.075 ms	15 - 19 MB	NPU
encoder	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	9.014 ms	16 - 333 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	7.232 ms	2 - 295 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X2 Elite	7.863 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® X Elite	19.138 ms	2 - 2 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	12.783 ms	2 - 349 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	47.835 ms	2 - 264 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	18.529 ms	2 - 3 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8775P	20.574 ms	2 - 264 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS9075	22.215 ms	2 - 12 MB	NPU
encoder	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	28.134 ms	0 - 331 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA7255P	47.835 ms	2 - 264 MB	NPU
encoder	QNN_DLC	float	Qualcomm® SA8295P	25.875 ms	2 - 266 MB	NPU
encoder	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	8.987 ms	2 - 304 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	4.428 ms	6 - 152 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	8.411 ms	6 - 228 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8275 (Proxy)	38.056 ms	7 - 172 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8550 (Proxy)	12.005 ms	7 - 9 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8775P	14.566 ms	7 - 175 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS9075	15.148 ms	6 - 66 MB	NPU
encoder	TFLITE	float	Qualcomm® QCS8450 (Proxy)	21.069 ms	7 - 341 MB	NPU
encoder	TFLITE	float	Qualcomm® SA7255P	38.056 ms	7 - 172 MB	NPU
encoder	TFLITE	float	Qualcomm® SA8295P	20.649 ms	7 - 288 MB	NPU
encoder	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	5.657 ms	5 - 154 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