Source record

Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2607.19907.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

TargetFinder is a computer vision system using fine-tuned YOLO models for real-time detection of GUI widgets across desktop platforms. Trained on 520 annotated screenshots (Windows, macOS, Ubuntu, web), it outperforms OmniParser and REMAUI baselines and enables system-wide deployment of target-aware pointing techniques like Bubble Cursor and Semantic Pointing. Dataset, models, and library are open-sourced.

Releases a cross-platform widget detection pipeline with dataset and models — a reusable building block for both accessibility pointing aids and agents that need to perceive arbitrary desktop GUIs.

Cited by

No citations on record.