EasyOCR Example

This document describes how to use the QAI AppBuilder Python API to run inference with the EasyOCR text recognition model on Qualcomm® Hexagon™ Processor (NPU).

Supported devices

Device	SoC
Fogwise® AIRbox Q900	QCS9075

Install QAI AppBuilder

tip

1. Install QAI AppBuilder by following the QAI AppBuilder installation guide.

2. Configure ADSP environment variables as described in Create ADSP environment variables.

Run the sample

Install dependencies

Install sample dependencies in the activated virtual environment:

Device

pip3 install requests tqdm qai-hub py3-wget Pillow torch torchvision opencv-python-headless easyocr

Run the script

• Enter the upstream samples directory

  Device

  cd qai-appbuilder/samples

• Prepare input data (bundled ch_en.png / english.png, or pass --Image_Path)

input image

warning

Verified on AIRbox Q900 (QCS9075): upstream easy_ocr.py uses DataLoader(..., num_workers=1), which forks worker processes and releases the QNN HTP context. Recognition then fails with:

RuntimeError: Cannot perform 'Inference': context 'EasyOCRRecognizer' has been released.

Before running, change every num_workers=int(workers) in Multimodal/Image_To_Text/easy_ocr/easy_ocr.py to num_workers=0 (or set the workers default in recognizer_get_text to 0).

• Run inference

  Device

  python3 Multimodal/Image_To_Text/easy_ocr/easy_ocr.py --chipset 9075

On success the terminal prints recognized Chinese and English lines, for example (default bilingual sample image):

白日依山尽
黄河入海流
欲穷千里目
更上一层楼
The sun beyond the mountain glows,
The Yellow River seawards flows_
You can enjoy a grander sight,
By climbing to a greater height。
Saving image to .../samples/build/image.png

tip

The first run downloads Detector / Recognizer models (including Chinese-English) via Qualcomm® AI Hub. Apply the num_workers=0 change first, then you can use the launcher:

python3 run_inference.py --list
python3 run_inference.py --model easy_ocr --args "--chipset 9075"