AI Agent Hub
Back to skills
Tencent Cloud Table Recognition V3 icon

Tencent Cloud Table Recognition V3

Office Efficiency Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please follow https://skillhub.cn/install/skillhub.md to install @tencent-adm/tencentcloud-ocr-recognizetableaccurate.

About this skill

Extracting table data from images or PDFs is a common need in office automation and data analysis. Traditional OCR tools often recognize scattered text but fail to preserve table structure, leading to messy extractions that require manual correction.

The Problem Solved

Processing documents with tables poses challenges: borderless tables are hard to detect, multiple or nested tables in one image cause recognition errors, and rotated images add complexity. Tencent Cloud Table Recognition V3 addresses these by accurately identifying various table types in Chinese/English images or PDFs and returning structured data.

Core Capabilities and Working Principle

The skill is based on Tencent Cloud OCR API, offering these capabilities:
- Regular table recognition: Accurately identifies tables with borders.
- Borderless table recognition: Detects and extracts content from tables without frames.
- Multi-table recognition: Identifies multiple independent tables in a single image.
- Nested table recognition: Handles complex cases where borderless tables are within bordered tables.
- Rotated table recognition: Supports table images at any rotation angle.
- Excel export: Converts recognition results directly into Excel files (Base64 encoded).
- PDF support: Enables single-page table recognition from PDF files.

Working principle: By calling the RecognizeTableAccurateOCR API endpoint, with input parameters like ImageBase64 (Base64-encoded image/PDF) or ImageUrl (URL of the image/PDF), it returns JSON output including Cells (list of cell texts, coordinates, and confidence) and TableCoordPoint (table vertex coordinates). Users can run it via scripts/main.py or directly use the tencentcloud-sdk-python library.

Usage Considerations

  • Environment: Requires Python 3.6+, installation of tencentcloud-sdk-python, and environment variables TENCENTCLOUD_SECRET_ID and TENCENTCLOUD_SECRET_KEY set.
  • Input limits: Encoded image/PDF must be under 10MB, resolution recommended at 600x800 or higher with aspect ratio less than 3. Supported formats: PNG, JPG, JPEG, BMP, PDF (single-page recognition).
  • API frequency: Default request limit is 2 calls per second; exceeding may trigger throttling.
  • Error handling: Common errors include FailedOperation.ImageDecodeFailed (image decoding failure), FailedOperation.ImageSizeTooLarge (oversized image), etc., requiring debugging based on error codes.
  • Best practices: Store images in Tencent Cloud COS for better download stability; prioritize high-resolution images for complex tables.

Use Cases

  • In finance departments, when processing scanned reimbursement form images, extract table data to automatically input into financial systems and reduce manual entry errors.
  • Legal teams receiving contract files in PDF format need to identify clause tables and export them to Excel for batch review and clause comparison.
  • Data analysts dealing with market survey images containing multiple tables need to convert them into structured Excel data for statistical analysis and visualization.
  • Office clerks processing order images with nested tables need to accurately recognize cell content and export to Excel for inventory record updates.

Best For

  • Financial auditors: Need to extract table data from scanned financial statements for verification and generating summary reports.
  • Data analysts: Need to convert tables from PDF reports into Excel files for data cleaning, modeling, and further analysis.
  • Office document managers: Need to process images with borderless or nested tables, accurately recognizing content to automate archiving workflows.
  • Sales operations specialists: Need to extract multi-table data from customer order images and consolidate into Excel for tracking sales performance.