---

name: easyocr description: "Use EasyOCR to detect and recognize text in images, configure model and language selection, load custom recognition bundles, and troubleshoot DBNet setup." metadata: disco-role: operating disable-model-invocation: true license: Apache 2.0

EasyOCR

Use this skill when the task is about extracting text from images, screenshots, scans, or cropped regions with EasyOCR.

Start here

1. Install EasyOCR together with a PyTorch build that matches your backend. For CPU-only use, a normal pip install easyocr is usually enough. For GPU runs, install the matching torch and torchvision wheels first.
2. Run scripts/inspect_runtime.py to confirm the package imports and the backend choice without downloading model weights.
3. Pick the right route:
   • sub-skills/inference/ for ordinary OCR API or CLI usage.
   • sub-skills/custom-models/ for custom recognition bundles or custom model directories.
   • sub-skills/dbnet/ for detect_network='dbnet18', DBNet import/init, or DCN compilation issues.

What this skill covers

• easyocr.Reader initialization and the main OCR methods.
• Language selection, model cache locations, download behavior, and backend fallback.
• The EasyOCR CLI entry point.
• Custom recognition model loading.
• DBNet detector setup and DCN operator compilation.

What this skill does not cover

• Training a new OCR model from scratch.
• ONNX export helpers.
• Maintainer-only language regeneration or release automation.

Quick import check

python -c "import easyocr; print(easyocr.__version__)"
python scripts/inspect_runtime.py --help

Route map

• Read references/api-reference.md for public signatures and output shapes.
• Read references/cli-reference.md for the CLI flag groups and parser caveats.
• Read references/configuration.md for environment variables, cache paths, backend selection, and model selection rules.
• Read references/troubleshooting.md for install, import, model-cache, and CLI/runtime quirks.
• Read references/repo-provenance.md to check the source commit, package version, and refresh baseline.
• Use sub-skills/inference/ for normal image OCR.
• Use sub-skills/custom-models/ for custom recognition bundles.
• Use sub-skills/dbnet/ for DBNet detector setup and DCN compilation.

Common user intents

• "Read text from this image" -> sub-skills/inference/
• "Use my own recognition model" -> sub-skills/custom-models/
• "Enable DBNet" or "compile DCN" -> sub-skills/dbnet/
• "EasyOCR import fails" -> references/troubleshooting.md

Notes for future agents

• Prefer the bundled scripts and references in this skill tree.
• Do not depend on the original repository checkout once the guidance has been distilled here.
• Treat easyocr.Reader and the CLI as the stable public surface; keep lower- level implementation details in the sub-skill references.