研究库 论文知识库

内容库 / 论文

arXiv:2609.19143 · 多模态

PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection

PANORAMA：基于掩码提议选择的全景接地描述生成

PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection

类型：arxiv
标识：2609.19143
链接：https://arxiv.org/abs/2609.19143
主分类：multimodal
形态：method
TLDR：Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded. Current vision-language models (VLMs) can generate fluent and detailed image captions, but reliably associating them with image pixels remains challenging. Existing methods that combine dense captioning with pixel-level grounding often produce either incomplete descriptions or inaccurate segmentation masks. We study this problem through panoptic grounded captioning, a task that requires a VLM to describe both foreground objects and background regions while grounding each ref
待LLM分类：否
标题中文：PANORAMA：基于掩码提议选择的全景接地描述生成
TLDR中文：在世界中行动的智能系统需要既全面又具备空间定位能力的图像理解。当前 vision-language models（VLM）能够生成流畅且详细的图像描述，但可靠地将描述与图像像素关联仍然具有挑战。现有将密集描述与像素级定位相结合的方法往往产生不完整的描述或不准确的分割掩码。我们通过全景接地描述生成任务研究此问题，该任务要求 VLM 同时描述前景物体与背景区域，并对每个引用...
来源文件：
/inbox/tom/_candidates/2026-09-17-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记 VaLR: Vision-aligned Latent Reasoning for Multi-modal LLM

笔记 flyP 短审稿 · Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

笔记 MMProLong:长上下文视觉语言模型的有效续训练(精读 · flyP)

笔记精读与批判：SenseNova-Vision — Vision as Unified Multimodal Generation

笔记精读：Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects

笔记 2026-07-04 上午多模态对照精读 · Seeker（文本→像素的长上下文 MLLM）

arXiv

2609.19143

TLDR

在世界中行动的智能系统需要既全面又具备空间定位能力的图像理解。当前 vision-language models（VLM）能够生成流畅且详细的图像描述，但可靠地将描述与图像像素关联仍然具有挑战。现有将密集描述与像素级定位相结合的方法往往产生不完整的描述或不准确的分割掩码。我们通过全景接地描述生成任务研究此问题，该任务要求 VLM 同时描述前景物体与背景区域，并对每个引用...Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded. Current vision-language models (VLMs) can generate fluent and detailed image captions, but reliably associating them with image pixels remains challenging. Existing methods that combine dense captioning with pixel-level grounding often produce either incomplete descriptions or inaccurate segmentation masks. We study this problem through panoptic grounded captioning, a task that requires a VLM to describe both foreground objects and background regions while grounding each ref

arXiv ↗ 中文全文