Implementation:Open compass VLMEvalKit ScreenSpotV2
| Field | Value |
|---|---|
| source | VLMEvalKit |
| domain | Vision, Benchmarking, GUI Grounding |
Overview
Benchmark dataset implementation for ScreenSpot V2 GUI element grounding evaluation in VLMEvalKit.
Description
ScreenSpotV2 inherits from ScreenSpot and extends the original ScreenSpot benchmark with updated evaluation logic for GUI agent grounding tasks. The TYPE field is set to 'GUI'. It reuses the base ScreenSpot infrastructure while providing V2-specific modifications for improved evaluation accuracy.
Usage
Registered in vlmeval/dataset/__init__.py and invoked through build_dataset() by benchmark name.
Code Reference
- Source:
vlmeval/dataset/GUI/screenspot_v2.py, Lines: L1-208 - Import:
from vlmeval.dataset.GUI.screenspot_v2 import ScreenSpotV2
Signature:
class ScreenSpotV2(ScreenSpot):
TYPE = "GUI"
...
I/O Contract
| Direction | Description |
|---|---|
| Inputs | TSV dataset file with screenshot images and GUI element grounding tasks |
| Outputs | Evaluation results DataFrame with accuracy scores per platform |
Usage Examples
from vlmeval.dataset import build_dataset
dataset = build_dataset('ScreenSpot_v2_Mobile')
Related Pages
Page Connections
Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment