---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-to-text

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/trocr/web-assets/model_demo.png)

# 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](https://huggingface.co/microsoft/trocr-small-stage1).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/trocr) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) 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](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/trocr/releases/v0.54.0/trocr-onnx-float.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/trocr/releases/v0.54.0/trocr-qnn_dlc-float.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/trocr/releases/v0.54.0/trocr-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[TrOCR on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/trocr)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/trocr) 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](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/trocr) 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 | 1.142 ms | 1 - 220 MB | NPU
| decoder | ONNX | float | Snapdragon® 8 Elite Mobile | 1.25 ms | 0 - 218 MB | NPU
| decoder | ONNX | float | Snapdragon® X2 Elite | 1.157 ms | 68 - 68 MB | NPU
| decoder | ONNX | float | Snapdragon® X Elite | 2.232 ms | 67 - 67 MB | NPU
| decoder | ONNX | float | Snapdragon® X Elite | 2.232 ms | 67 - 67 MB | NPU
| decoder | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 1.435 ms | 0 - 247 MB | NPU
| decoder | ONNX | float | Qualcomm® QCS8550 (Proxy) | 2.101 ms | 0 - 74 MB | NPU
| decoder | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.25 ms | 0 - 218 MB | NPU
| decoder | ONNX | float | Qualcomm® QCS9075 | 2.739 ms | 7 - 16 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.14 ms | 1 - 172 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 1.237 ms | 0 - 186 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® X2 Elite | 1.598 ms | 7 - 7 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® X Elite | 2.131 ms | 7 - 7 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® X Elite | 2.131 ms | 7 - 7 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.389 ms | 0 - 276 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 1.95 ms | 3 - 4 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® SA8775P | 2.873 ms | 7 - 106 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® SA8775P | 2.873 ms | 7 - 106 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® SA8775P | 2.873 ms | 7 - 106 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® SA7255P | 4.136 ms | 7 - 105 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® SA8295P | 2.627 ms | 0 - 49 MB | NPU
| decoder | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.237 ms | 0 - 186 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® QCS9075 | 2.525 ms | 7 - 15 MB | NPU
| decoder | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 2.683 ms | 2 - 217 MB | NPU
| decoder | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.116 ms | 0 - 175 MB | NPU
| decoder | TFLITE | float | Snapdragon® 8 Elite Mobile | 1.214 ms | 0 - 190 MB | NPU
| decoder | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.404 ms | 0 - 281 MB | NPU
| decoder | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 2.042 ms | 0 - 2 MB | NPU
| decoder | TFLITE | float | Qualcomm® SA8775P | 2.876 ms | 0 - 102 MB | NPU
| decoder | TFLITE | float | Qualcomm® SA8775P | 2.876 ms | 0 - 102 MB | NPU
| decoder | TFLITE | float | Qualcomm® SA8775P | 2.876 ms | 0 - 102 MB | NPU
| decoder | TFLITE | float | Qualcomm® SA7255P | 4.202 ms | 0 - 103 MB | NPU
| decoder | TFLITE | float | Qualcomm® SA8295P | 2.669 ms | 0 - 43 MB | NPU
| decoder | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 1.214 ms | 0 - 190 MB | NPU
| decoder | TFLITE | float | Qualcomm® QCS9075 | 2.594 ms | 0 - 83 MB | NPU
| decoder | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 2.459 ms | 0 - 217 MB | NPU
| encoder | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.156 ms | 1 - 150 MB | NPU
| encoder | ONNX | float | Snapdragon® 8 Elite Mobile | 5.278 ms | 16 - 166 MB | NPU
| encoder | ONNX | float | Snapdragon® X2 Elite | 4.579 ms | 48 - 48 MB | NPU
| encoder | ONNX | float | Snapdragon® X Elite | 11.423 ms | 47 - 47 MB | NPU
| encoder | ONNX | float | Snapdragon® X Elite | 11.423 ms | 47 - 47 MB | NPU
| encoder | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 7.824 ms | 16 - 234 MB | NPU
| encoder | ONNX | float | Qualcomm® QCS8550 (Proxy) | 10.913 ms | 0 - 60 MB | NPU
| encoder | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 5.278 ms | 16 - 166 MB | NPU
| encoder | ONNX | float | Qualcomm® QCS9075 | 14.132 ms | 15 - 19 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.31 ms | 2 - 115 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 5.447 ms | 2 - 114 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® X2 Elite | 5.055 ms | 2 - 2 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® X Elite | 11.986 ms | 2 - 2 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® X Elite | 11.986 ms | 2 - 2 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 8.006 ms | 0 - 189 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 11.264 ms | 2 - 4 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® SA8775P | 13.992 ms | 2 - 108 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® SA8775P | 13.992 ms | 2 - 108 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® SA8775P | 13.992 ms | 2 - 108 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® SA7255P | 37.067 ms | 2 - 106 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® SA8295P | 19.685 ms | 2 - 233 MB | NPU
| encoder | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 5.447 ms | 2 - 114 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® QCS9075 | 15.054 ms | 2 - 12 MB | NPU
| encoder | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 19.555 ms | 0 - 307 MB | NPU
| encoder | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.074 ms | 0 - 110 MB | NPU
| encoder | TFLITE | float | Snapdragon® 8 Elite Mobile | 5.205 ms | 6 - 118 MB | NPU
| encoder | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 7.714 ms | 7 - 198 MB | NPU
| encoder | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 10.695 ms | 7 - 9 MB | NPU
| encoder | TFLITE | float | Qualcomm® SA8775P | 13.51 ms | 7 - 111 MB | NPU
| encoder | TFLITE | float | Qualcomm® SA8775P | 13.51 ms | 7 - 111 MB | NPU
| encoder | TFLITE | float | Qualcomm® SA8775P | 13.51 ms | 7 - 111 MB | NPU
| encoder | TFLITE | float | Qualcomm® SA7255P | 36.817 ms | 7 - 120 MB | NPU
| encoder | TFLITE | float | Qualcomm® SA8295P | 19.074 ms | 8 - 236 MB | NPU
| encoder | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 5.205 ms | 6 - 118 MB | NPU
| encoder | TFLITE | float | Qualcomm® QCS9075 | 13.937 ms | 6 - 65 MB | NPU
| encoder | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 19.779 ms | 7 - 311 MB | NPU

## License
* The license for the original implementation of TrOCR can be found
  [here](https://github.com/microsoft/unilm/blob/master/LICENSE).

## References
* [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282)
* [Source Model Implementation](https://huggingface.co/microsoft/trocr-small-stage1)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).