Guardado en:
Detalles Bibliográficos
Autores principales: Kumar, Ashutosh, Saini, Rajat, Pan, Jingjing, Erdogan, Mustafa, Zhang, Mingfang, Dem, Betty Le, Kobori, Norimasa, Kong, Quan
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2604.08337
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908949575368704
author Kumar, Ashutosh
Saini, Rajat
Pan, Jingjing
Erdogan, Mustafa
Zhang, Mingfang
Dem, Betty Le
Kobori, Norimasa
Kong, Quan
author_facet Kumar, Ashutosh
Saini, Rajat
Pan, Jingjing
Erdogan, Mustafa
Zhang, Mingfang
Dem, Betty Le
Kobori, Norimasa
Kong, Quan
contents Current vision-language pre-training (VLP) paradigms excel at global scene understanding but struggle with instance-level reasoning due to global-only supervision. We introduce InstAP, an Instance-Aware Pre-training framework that jointly optimizes global vision-text alignment and fine-grained, instance-level contrastive alignment by grounding textual mentions to specific spatial-temporal regions. To support this, we present InstVL, a large-scale dataset (2 million images, 50,000 videos) with dual-granularity annotations: holistic scene captions and dense, grounded instance descriptions. On the InstVL benchmark, InstAP substantially outperforms existing VLP models on instance-level retrieval, and also surpasses a strong VLP baseline trained on the exact same data corpus, isolating the benefit of our instance-aware objective. Moreover, instance-centric pre-training improves global understanding: InstAP achieves competitive zero-shot performance on multiple video benchmarks, including MSR-VTT and DiDeMo. Qualitative visualizations further show that InstAP localizes textual mentions to the correct instances, while global-only models exhibit more diffuse, scene-level attention.
format Preprint
id arxiv_https___arxiv_org_abs_2604_08337
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle InstAP: Instance-Aware Vision-Language Pre-Train for Spatial-Temporal Understanding
Kumar, Ashutosh
Saini, Rajat
Pan, Jingjing
Erdogan, Mustafa
Zhang, Mingfang
Dem, Betty Le
Kobori, Norimasa
Kong, Quan
Computer Vision and Pattern Recognition
Artificial Intelligence
Current vision-language pre-training (VLP) paradigms excel at global scene understanding but struggle with instance-level reasoning due to global-only supervision. We introduce InstAP, an Instance-Aware Pre-training framework that jointly optimizes global vision-text alignment and fine-grained, instance-level contrastive alignment by grounding textual mentions to specific spatial-temporal regions. To support this, we present InstVL, a large-scale dataset (2 million images, 50,000 videos) with dual-granularity annotations: holistic scene captions and dense, grounded instance descriptions. On the InstVL benchmark, InstAP substantially outperforms existing VLP models on instance-level retrieval, and also surpasses a strong VLP baseline trained on the exact same data corpus, isolating the benefit of our instance-aware objective. Moreover, instance-centric pre-training improves global understanding: InstAP achieves competitive zero-shot performance on multiple video benchmarks, including MSR-VTT and DiDeMo. Qualitative visualizations further show that InstAP localizes textual mentions to the correct instances, while global-only models exhibit more diffuse, scene-level attention.
title InstAP: Instance-Aware Vision-Language Pre-Train for Spatial-Temporal Understanding
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2604.08337