LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding

Hugging Face Daily Papers Papers
vision-language visual-grounding object-detection parallel-decoding speed-accuracy large-scale-dataset

Summary

LocateAnything proposes Parallel Box Decoding for unified visual grounding and object detection, decoding geometric elements as atomic units to improve throughput and localization accuracy, supported by a large-scale dataset of 138M samples.

Vision-language models (VLMs) commonly formulate visual grounding and detection as a coordinate-token generation problem, serializing each 2D box into multiple 1D tokens that are learned and decoded largely independently. This token-by-token decoding mismatches the coupled structure of box geometry and creates a practical inference bottleneck due to strictly sequential generation. We introduce LocateAnything, a unified generative grounding and detection framework based on Parallel Box Decoding (PBD). By decoding geometric elements such as bounding boxes and points as atomic units in a single step, LocateAnything preserves intra-box geometric coherence and unlocks substantial parallelism. We show that PBD improves both decoding throughput and localization accuracy. We further develop a scalable data engine and curate LocateAnything-Data, a large-scale dataset with more than 138 million training samples, substantially increasing data diversity for high-precision localization. Extensive evaluations show that LocateAnything advances the speed-accuracy frontier, achieving significantly higher decoding throughput while improving high-IoU localization quality across diverse benchmarks. The results highlight the complementary benefits of Parallel Box Decoding and large-scale training data in enabling efficient and precise unified visual grounding and detection.
Original Article
View Cached Full Text

Cached at: 05/27/26, 02:47 AM

Paper page - LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding

Source: https://huggingface.co/papers/2605.27365 Authors:

,

,

,

,

,

,

,

,

,

,

,

Abstract

Parallel Box Decoding enables efficient and accurate unified visual grounding and detection by decoding geometric elements as atomic units, improving both throughput and localization quality.

Vision-language models(VLMs) commonly formulatevisual groundinganddetectionas acoordinate-token generationproblem, serializing each 2D box into multiple 1D tokens that are learned and decoded largely independently. This token-by-token decoding mismatches the coupled structure ofbox geometryand creates a practical inference bottleneck due to strictly sequential generation. We introduce LocateAnything, a unified generative grounding anddetectionframework based onParallel Box Decoding(PBD). By decoding geometric elements such as bounding boxes and points as atomic units in a single step, LocateAnything preserves intra-boxgeometric coherenceand unlocks substantial parallelism. We show that PBD improves bothdecoding throughputandlocalization accuracy. We further develop a scalable data engine and curate LocateAnything-Data, a large-scale dataset with more than 138 million training samples, substantially increasing data diversity for high-precision localization. Extensive evaluations show that LocateAnything advances the speed-accuracy frontier, achieving significantly higherdecoding throughputwhile improving high-IoU localization quality across diverse benchmarks. The results highlight the complementary benefits ofParallel Box Decodingandlarge-scale training datain enabling efficient and precise unifiedvisual groundinganddetection.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2605\.27365

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.27365 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.27365 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.27365 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

@VincentLogic: NVIDIA's newly open-sourced LocateAnything model is really impressive. The previous visual grounding models generated coordinates digit by digit (like squeezing toothpaste), slow and unstable. This new model uses "parallel bounding box decoding" to predict complete coordinates in one step, much faster and more accurate...

X AI KOLs Timeline

NVIDIA has open-sourced the LocateAnything model, using parallel bounding box decoding technology to predict complete coordinates in one step, fast and accurate. The model has only 3B parameters and can run on consumer-grade GPUs, supporting video object localization, UI recognition, OCR, and other tasks.

Vision-Language Grounding as Bidirectional Concept Correspondence

Hugging Face Daily Papers

This paper introduces ConCor-1, a grounding model that treats vision-language grounding as bidirectional concept correspondence, jointly recovering text spans, image segments, and cross-modal matches without prespecified phrases. It unifies phrase grounding, referring expression grounding, and open-vocabulary detection, achieving significant F1 improvements on long-caption and zero-shot LVIS benchmarks.