--- 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/easyocr/web-assets/model_demo.png) # EasyOCR: Optimized for Qualcomm Devices EasyOCR is a machine learning model that can recognize text in images. It supports 80+ supported languages and all popular writing scripts. This is based on the implementation of EasyOCR found [here](https://github.com/JaidedAI/EasyOCR). 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/easyocr) 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/easyocr/releases/v0.54.0/easyocr-onnx-float.zip) | ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/easyocr/releases/v0.54.0/easyocr-onnx-w8a8.zip) | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/easyocr/releases/v0.54.0/easyocr-tflite-float.zip) | TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/easyocr/releases/v0.54.0/easyocr-tflite-w8a8.zip) For more device-specific assets and performance metrics, visit **[EasyOCR on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/easyocr)**. ### 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/easyocr) 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 [EasyOCR on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/easyocr) for usage instructions. ## Model Details **Model Type:** Model_use_case.image_to_text **Model Stats:** - Model checkpoint: easyocr-small-stage1 - Input resolution: 608x800 - Number of parameters (detector): 20.8M - Model size (detector) (float): 79.2 MB - Number of parameters (recognizer): 3.84M - Model size (recognizer) (float): 14.7 MB ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | detector | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 16.117 ms | 7 - 176 MB | NPU | detector | ONNX | float | Snapdragon® 8 Elite Mobile | 21.877 ms | 3 - 171 MB | NPU | detector | ONNX | float | Snapdragon® X2 Elite | 20.02 ms | 36 - 36 MB | NPU | detector | ONNX | float | Snapdragon® X Elite | 38.281 ms | 35 - 35 MB | NPU | detector | ONNX | float | Snapdragon® X Elite | 38.281 ms | 35 - 35 MB | NPU | detector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 27.179 ms | 3 - 237 MB | NPU | detector | ONNX | float | Qualcomm® QCS8550 (Proxy) | 37.381 ms | 6 - 8 MB | NPU | detector | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 21.877 ms | 3 - 171 MB | NPU | detector | ONNX | float | Qualcomm® QCS9075 | 69.962 ms | 5 - 14 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.358 ms | 0 - 162 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 7.762 ms | 2 - 159 MB | NPU | detector | ONNX | w8a8 | Snapdragon® X2 Elite | 6.717 ms | 20 - 20 MB | NPU | detector | ONNX | w8a8 | Snapdragon® X Elite | 13.655 ms | 19 - 19 MB | NPU | detector | ONNX | w8a8 | Snapdragon® X Elite | 13.655 ms | 19 - 19 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 9.741 ms | 2 - 218 MB | NPU | detector | ONNX | w8a8 | Qualcomm® QCS6490 | 955.669 ms | 229 - 248 MB | CPU | detector | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 13.068 ms | 1 - 4 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 917.984 ms | 234 - 241 MB | CPU | detector | ONNX | w8a8 | Qualcomm® QCM6690 | 929.421 ms | 231 - 238 MB | CPU | detector | ONNX | w8a8 | Qualcomm® QCS9075 | 13.593 ms | 1 - 4 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 7.762 ms | 2 - 159 MB | NPU | detector | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 917.984 ms | 234 - 241 MB | CPU | detector | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 16.549 ms | 6 - 183 MB | NPU | detector | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 21.933 ms | 6 - 170 MB | NPU | detector | QNN_DLC | float | Snapdragon® X2 Elite | 20.718 ms | 6 - 6 MB | NPU | detector | QNN_DLC | float | Snapdragon® X Elite | 39.398 ms | 6 - 6 MB | NPU | detector | QNN_DLC | float | Snapdragon® X Elite | 39.398 ms | 6 - 6 MB | NPU | detector | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 28.043 ms | 6 - 236 MB | NPU | detector | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 37.784 ms | 6 - 37 MB | NPU | detector | QNN_DLC | float | Qualcomm® SA8775P | 67.953 ms | 1 - 171 MB | NPU | detector | QNN_DLC | float | Qualcomm® SA8775P | 67.953 ms | 1 - 171 MB | NPU | detector | QNN_DLC | float | Qualcomm® SA8775P | 67.953 ms | 1 - 171 MB | NPU | detector | QNN_DLC | float | Qualcomm® SA7255P | 269.364 ms | 1 - 171 MB | NPU | detector | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 80.636 ms | 6 - 258 MB | NPU | detector | QNN_DLC | float | Qualcomm® SA8295P | 77.649 ms | 0 - 183 MB | NPU | detector | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 21.933 ms | 6 - 170 MB | NPU | detector | QNN_DLC | float | Qualcomm® QCS9075 | 75.907 ms | 6 - 14 MB | NPU | detector | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 16.479 ms | 0 - 180 MB | NPU | detector | TFLITE | float | Snapdragon® 8 Elite Mobile | 21.855 ms | 1 - 169 MB | NPU | detector | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 27.994 ms | 1 - 259 MB | NPU | detector | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 38.201 ms | 1 - 3 MB | NPU | detector | TFLITE | float | Qualcomm® SA8775P | 67.995 ms | 1 - 175 MB | NPU | detector | TFLITE | float | Qualcomm® SA8775P | 67.995 ms | 1 - 175 MB | NPU | detector | TFLITE | float | Qualcomm® SA8775P | 67.995 ms | 1 - 175 MB | NPU | detector | TFLITE | float | Qualcomm® SA7255P | 269.383 ms | 1 - 174 MB | NPU | detector | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 80.637 ms | 1 - 283 MB | NPU | detector | TFLITE | float | Qualcomm® SA8295P | 73.897 ms | 1 - 184 MB | NPU | detector | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 21.855 ms | 1 - 169 MB | NPU | detector | TFLITE | float | Qualcomm® QCS9075 | 75.431 ms | 0 - 49 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 4.459 ms | 0 - 159 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 5.759 ms | 0 - 159 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.171 ms | 0 - 209 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® QCS6490 | 51.313 ms | 0 - 24 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 9.486 ms | 0 - 34 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® SA8775P | 10.044 ms | 0 - 157 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® SA8775P | 10.044 ms | 0 - 157 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® SA8775P | 10.044 ms | 0 - 157 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 18.728 ms | 0 - 173 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® QCM6690 | 243.364 ms | 0 - 209 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® QCS9075 | 11.825 ms | 0 - 24 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 5.759 ms | 0 - 159 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 14.689 ms | 0 - 211 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® SA7255P | 35.802 ms | 0 - 155 MB | NPU | detector | TFLITE | w8a8 | Qualcomm® SA8295P | 19.013 ms | 0 - 153 MB | NPU | detector | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 18.728 ms | 0 - 173 MB | NPU | recognizer | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.486 ms | 0 - 791 MB | NPU | recognizer | ONNX | float | Snapdragon® 8 Elite Mobile | 10.234 ms | 0 - 639 MB | NPU | recognizer | ONNX | float | Snapdragon® X2 Elite | 11.896 ms | 13 - 13 MB | NPU | recognizer | ONNX | float | Snapdragon® X Elite | 16.238 ms | 11 - 11 MB | NPU | recognizer | ONNX | float | Snapdragon® X Elite | 16.238 ms | 11 - 11 MB | NPU | recognizer | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 12.088 ms | 0 - 764 MB | NPU | recognizer | ONNX | float | Qualcomm® QCS8550 (Proxy) | 17.032 ms | 0 - 31 MB | NPU | recognizer | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.234 ms | 0 - 639 MB | NPU | recognizer | ONNX | float | Qualcomm® QCS9075 | 18.818 ms | 0 - 3 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 9.268 ms | 1 - 873 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 8.306 ms | 1 - 701 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® X2 Elite | 10.77 ms | 14 - 14 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® X Elite | 12.299 ms | 10 - 10 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® X Elite | 12.299 ms | 10 - 10 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 8.996 ms | 2 - 883 MB | NPU | recognizer | ONNX | w8a8 | Qualcomm® QCS6490 | 122.172 ms | 20 - 24 MB | CPU | recognizer | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 11.978 ms | 0 - 440 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 40.951 ms | 17 - 24 MB | CPU | recognizer | ONNX | w8a8 | Qualcomm® QCM6690 | 55.882 ms | 19 - 25 MB | CPU | recognizer | ONNX | w8a8 | Qualcomm® QCS9075 | 14.86 ms | 2 - 5 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 8.306 ms | 1 - 701 MB | NPU | recognizer | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 40.951 ms | 17 - 24 MB | CPU | recognizer | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.416 ms | 0 - 734 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 10.27 ms | 0 - 697 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® X2 Elite | 12.242 ms | 0 - 0 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® X Elite | 16.202 ms | 0 - 0 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® X Elite | 16.202 ms | 0 - 0 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 12.43 ms | 0 - 831 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 17.232 ms | 0 - 3 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® SA8775P | 20.327 ms | 0 - 700 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® SA8775P | 20.327 ms | 0 - 700 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® SA8775P | 20.327 ms | 0 - 700 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® SA7255P | 47.52 ms | 0 - 692 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 22.296 ms | 0 - 403 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® SA8295P | 24.778 ms | 0 - 338 MB | NPU | recognizer | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.27 ms | 0 - 697 MB | NPU | recognizer | QNN_DLC | float | Qualcomm® QCS9075 | 18.909 ms | 0 - 2 MB | NPU | recognizer | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 39.498 ms | 10 - 21 MB | CPU | recognizer | TFLITE | float | Snapdragon® 8 Elite Mobile | 45.034 ms | 7 - 15 MB | CPU | recognizer | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 52.047 ms | 6 - 15 MB | CPU | recognizer | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 70.537 ms | 6 - 8 MB | CPU | recognizer | TFLITE | float | Qualcomm® SA8775P | 99.522 ms | 9 - 14 MB | CPU | recognizer | TFLITE | float | Qualcomm® SA8775P | 99.522 ms | 9 - 14 MB | CPU | recognizer | TFLITE | float | Qualcomm® SA8775P | 99.522 ms | 9 - 14 MB | CPU | recognizer | TFLITE | float | Qualcomm® SA7255P | 173.475 ms | 8 - 16 MB | CPU | recognizer | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 76.956 ms | 6 - 16 MB | CPU | recognizer | TFLITE | float | Qualcomm® SA8295P | 68.884 ms | 5 - 10 MB | CPU | recognizer | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 45.034 ms | 7 - 15 MB | CPU | recognizer | TFLITE | float | Qualcomm® QCS9075 | 135.05 ms | 4 - 33 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 26.959 ms | 9 - 19 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 29.368 ms | 7 - 19 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 32.709 ms | 5 - 15 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® QCS6490 | 141.598 ms | 3 - 22 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 36.767 ms | 5 - 6 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® SA8775P | 70.718 ms | 7 - 13 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® SA8775P | 70.718 ms | 7 - 13 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® SA8775P | 70.718 ms | 7 - 13 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 56.757 ms | 8 - 15 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® QCM6690 | 76.78 ms | 6 - 13 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® QCS9075 | 98.783 ms | 3 - 21 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 29.368 ms | 7 - 19 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 46.053 ms | 6 - 17 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® SA7255P | 140.038 ms | 8 - 17 MB | CPU | recognizer | TFLITE | w8a8 | Qualcomm® SA8295P | 50.717 ms | 6 - 11 MB | CPU | recognizer | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 56.757 ms | 8 - 15 MB | CPU ## License * The license for the original implementation of EasyOCR can be found [here](https://github.com/JaidedAI/EasyOCR/blob/master/LICENSE). ## References * [Source Model Implementation](https://github.com/JaidedAI/EasyOCR) ## 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).