Implementation:Open compass VLMEvalKit CCOCRDataset

Field	Value
source	VLMEvalKit
domain	Vision, Benchmarking, OCR, Document Understanding

Overview

Benchmark dataset implementation for CC-OCR comprehensive OCR evaluation in VLMEvalKit.

Description

CCOCRDataset inherits from ImageBaseDataset and implements the CC-OCR benchmark for evaluating OCR capabilities across multiple tasks including document parsing, key information extraction (KIE), and multi-language OCR. The TYPE field is set to 'VQA'. It provides extensive dataset URLs from ModelScope covering document photos/scans in Chinese/English, table parsing, molecular/formula recognition, KIE tasks (SROIE, CORD, EPHOIE), and multi-language OCR for 8+ languages.

Usage

Registered in vlmeval/dataset/__init__.py and invoked through build_dataset() by benchmark name.

Code Reference

Source: vlmeval/dataset/image_ccocr.py, Lines: L1-303
Import: from vlmeval.dataset.image_ccocr import CCOCRDataset

Signature:

class CCOCRDataset(ImageBaseDataset):
    TYPE = 'VQA'
    DATASET_URL_MODELSCOPE = {...}
    ...

I/O Contract

Direction	Description
Inputs	TSV dataset file with document/text images and OCR tasks
Outputs	Evaluation results DataFrame with OCR accuracy scores per task type

Usage Examples

from vlmeval.dataset import build_dataset
dataset = build_dataset('CCOCR_DocParsing_DocPhotoChn')

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment