Implementation:Open compass VLMEvalKit ScreenSpot Pro

Field	Value
source	VLMEvalKit
domain	Vision, Benchmarking, GUI Grounding, Professional Software

Overview

Benchmark dataset implementation for ScreenSpot Pro professional software GUI grounding evaluation in VLMEvalKit.

Description

ScreenSpot_Pro inherits from ImageBaseDataset and implements the ScreenSpot Pro benchmark for evaluating GUI agent capabilities on professional software interfaces. The TYPE field is set to 'GUI'. It covers six professional domains: Development, Creative, CAD, Scientific, Office, and OS, with point-based evaluation and functional referring expressions.

Usage

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

Code Reference

Source: vlmeval/dataset/GUI/screenspot_pro.py, Lines: L1-460
Import: from vlmeval.dataset.GUI.screenspot_pro import ScreenSpot_Pro

Signature:

class ScreenSpot_Pro(ImageBaseDataset):
    MODALITY = "IMAGE"
    TYPE = "GUI"
    DATASET_URL = {...}
    DATASET_MD5 = {...}
    EVAL_TYPE = "point"
    RE_TYPE = "functional"
    ...

I/O Contract

Direction	Description
Inputs	TSV dataset file with professional software screenshots and element grounding tasks
Outputs	Evaluation results DataFrame with accuracy scores per professional domain

Usage Examples

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

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment