Implementation:Open compass VLMEvalKit CGAVCounting

Field	Value
source	VLMEvalKit
domain	Vision, Benchmarking, Video Counting

Overview

Benchmark dataset implementation for CG-AV-Counting video counting evaluation in VLMEvalKit.

Description

CGAVCounting inherits from VideoBaseDataset and implements the CG-AV-Counting benchmark for audio-visual counting tasks in video. The TYPE field is set to 'Video-Counting'. It supports frame-based clue extraction with configurable frame time usage and provides custom video frame sampling based on timestamps.

Usage

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

Code Reference

Source: vlmeval/dataset/CGAVCounting/cg_av_counting.py, Lines: L1-405
Import: from vlmeval.dataset.CGAVCounting.cg_av_counting import CGAVCounting

Signature:

class CGAVCounting(VideoBaseDataset):
    TYPE = "Video-Counting"
    dataset = "CG-AV-Counting"
    MD5 = "d1cd8486353ab85178098d443264a7d0"
    ...

I/O Contract

Direction	Description
Inputs	TSV dataset file with video paths and counting questions
Outputs	Evaluation results DataFrame with scores per category

Usage Examples

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

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment