LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding
2026-05-26 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine LearningRobotics
AI summaryⓘ
The authors propose LocateAnything, a new method for detecting objects and locating regions in images using vision-language models. Instead of predicting box coordinates step-by-step, their Parallel Box Decoding (PBD) predicts all parts of a box at once, making the process faster and more accurate. They also created a huge dataset with over 138 million examples to help train their system better. Tests show that their approach improves both speed and precision compared to previous methods. Overall, the authors demonstrate that decoding all box elements simultaneously and using a large dataset helps improve object detection and localization.
vision-language modelsvisual groundingobject detectionbounding boxparallel decodinglocalization accuracyinference throughputlarge-scale datasethigh-IoUgenerative models
Authors
Shihao Wang, Shilong Liu, Yuanguo Kuang, Xinyu Wei, Yangzhou Liu, Zhiqi Li, Yunze Man, Guo Chen, Andrew Tao, Guilin Liu, Jan Kautz, Lei Zhang, Zhiding Yu
Abstract
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.