DOCSEARCHbyAlgolia
Submit Docs

EdgeOCR Documentation

website·Indexed 30 days ago

https://guide.edge-ocr.net/

Developer and user guides for EdgeOCR, featuring an SDK for OCR and barcode recognition with APIs like EdgeVisionAPI, NefrockLicenseAPI, and YotsukaD API. Includes documentation for EdgeOCR Matching, a specialized tool for comparing scanned data from source documents and actual products, covering flow configuration, task settings, and worker operation guides.

Tokens
44.4K
Snippets
87
Records
356
Agent score
49%

What's inside EdgeOCR

  1. Overview of EdgeOCR value conversion settings

    Conversion settings are used to process read values before matching to absorb notation discrepancies (e.g., different code systems, extra characters, or sequential numbering).

    Conversion settings can be applied in the following locations:

    • Matching Source Settings: In 'Import Settings' (per read item).
    • Matching Source Settings: In 'Additional Conversions' (per item).
    • Matching Destination Settings: In 'Additional Conversions' (per destination read item).

    Each conversion consists of:

    • Item Name: The name assigned to the value after conversion. This name is used in item settings and matching destination mapping.
    • Conversion Type: Selection of 'None', 'Master Conversion', 'Delimiter', 'Regular Expression', 'Range Expansion', or 'Position Specification'.
    • Additional Conversion (Toggle): When ON, allows layering another conversion on top of the current result via 'Additional Conversion Settings'.

    Note: If a conversion fails (e.g., no matching value in a Master Conversion), the reading screen will display "読み取り結果の変換に失敗しました" (Failed to convert reading result).

  2. Overview of EdgeOCR Matching capabilities

    EdgeOCR Matching is a built-in feature of the EdgeOCR APP used to verify if two items (e.g., a slip/instruction sheet and a product label) match by scanning text or barcodes via camera. It is designed to streamline tasks such as shipping inspection, receiving inspection, and picking inspections.

  4. Understand the Matching Flow screen layout

    The Matching Flow screen is the starting point for matching operations after selecting a flow from the flow list. It contains the following key sections:

    • Header Information: Displays input fields if Header Settings are enabled. Users can tap the camera icon to scan (e.g., order numbers) or enter data manually. Header content is included in the matching result logs.
    • Total Quantity: Displays if Overall Quantity Settings are enabled. Users register the total quantity via scanning or manual input. Progress is shown as 'Current Total Quantity / Total Quantity'.
    • Instruction Comments: Displays guidance text for the worker if configured in the Flow Detail Settings.
    • Task List: Shows the matching status of each task in the flow as cards. Each card displays the source reading and, if quantity settings are enabled, 'Quantity: Current / Required'. Tapping a card opens the Task Detail screen. If a source reading is duplicated across tasks, the card is highlighted with a warning: '※ 重複しているタスクがあります' (Duplicate tasks exist).

  5. Understand the EdgeOCR Matching App configuration structure

    The EdgeOCR Matching App is organized into a hierarchy of four elements to manage business processes:

    • Flow: A collection of settings corresponding to a single business operation (e.g., "Shipping Inspection").
    • Task: A unit of matching work executed sequentially within a Flow.
    • Source (照合元): The reading target used as the reference standard (e.g., a slip or instruction sheet).
    • Destination (照合先): The reading target compared against the Source (e.g., a product label).

    The basic operational flow is: Read Source $\rightarrow$ Read Destination $\rightarrow$ Validate match (Success if identical, Error if mismatched).

  6. Understand EdgeOCR image analysis and aspect ratio handling

    The area analyzed by the scan method depends on the model used. For example, model-d320x320 expects a 1:1 aspect ratio (320x320). If the input image has a different aspect ratio, the SDK crops the image (defaulting to the center) to match the model's aspect ratio, using the input image's width as the base.

    To improve detection accuracy, align the input image's aspect ratio with the model's input size to avoid padding. For model-d320x320, an input image close to a 1:1 ratio is most effective. If a narrow crop (e.g., 1080x480) is used with a 1:1 model, the SDK will add black padding to the vertical axis, and these padded areas will not be analyzed.

  7. EdgeOCR Documentation Overview

    The EdgeOCR documentation serves as a comprehensive manual for both first-time field operators and administrators/leaders supporting internal deployment. It provides step-by-step guidance from initial setup to daily operations, regardless of the user's technical expertise.

  8. Understand the Flow and Task hierarchy in EdgeOCR

    The matching application organizes work using a two-tier hierarchy: Flows and Tasks.

    • Flow (フロー): A collection of matching settings corresponding to a single business unit (e.g., 'Shipping Inspection' or 'Receiving Inspection'). Workers select a Flow from the list to begin their work. A single Flow can contain multiple Tasks.
    • Task (タスク): The individual unit of matching work executed within a Flow. Each Task defines the source, destination, and quantity settings, and they are executed in the order they were registered.

  9. Configure the Matching Source for EdgeOCR tasks

    The Matching Source setting defines the baseline reading (such as an invoice, instruction sheet, or picking list) used for comparison. This is configured in the 'Matching Source Settings' (照合元設定) section of the Task Settings screen. The data flow follows this sequence: Scenario Reading → Intake Settings (Conversion) → Item Settings (Concatenation) → Additional Conversion → Final Value used for matching.

  10. Configure Matching Destination (照合先設定) for OCR tasks

    The Matching Destination settings define the reading process for items (such as product labels or packaging) that will be compared against a matching source. These settings are accessed via the 'Matching Destination Settings' (照合先設定) in the task configuration screen.

  11. Understand the EdgeOCR APP core concepts and mental model

    EdgeOCR APP extracts specific strings, numbers, and barcodes from camera images to assist in field operations. It uses a modular configuration system called 'Scenarios' to ensure consistent reading procedures based on the specific goal of the task. The system is built on four primary elements:

    • Scan Definition (What to read): The reading rules, including the type of target (e.g., numeric, date, barcode, dictionary registration) and validation rules (digit count, format, excluded characters).
    • Reading Mode (How to proceed): Determines the sequence of capture, such as Single Read (one by one), Simultaneous Read (multiple items at once), or Continuous Read (sequential capture).
    • Output Mode (Where to send data): Defines the destination of the extracted data, such as the Clipboard, File Output, or Direct Input mode (sending data directly to another application).
    • Scenario (Task-specific profile): A complete execution profile that bundles a Scan Definition, Reading Mode, Output Mode, and Input Device together for a specific field objective.

  12. Understand Scan Definitions and Scenarios in EdgeOCR

    In EdgeOCR, reading operations are managed through two primary concepts:

    • Scan Definition: Defines what to read. It specifies the target data format, such as using regular expressions to isolate specific patterns (e.g., item codes).
    • Scenario: An execution profile that defines how to read. It bundles a scan mode (e.g., single read) with a specific scan definition. Matching applications reference these scenarios to perform OCR tasks.