文章目录
- 0. Abstract
- 1. Main
- Reference
- 彩蛋:katex
0. Abstract
概念性总结,机器翻译不准,以原文为准
大多数现有的方法都是免费的开放词汇
Most existing methods for training-free Open-Vocabulary
语义分割(OVSS)是基于CLIP的。
Semantic Segmentation (OVSS) are based on CLIP.
而
While
这些方法虽然取得了进展,但往往面临挑战
these approaches have made progress, they often face chal-
在精确定位或需要复杂的管道长度
lenges in precise localization or require complex pipelines
把单独的模块组合起来,尤指在遥感中
to combine separate modules, especially in remote sens-
在目前众多密集和小目标的场景中
ing scenarios where numerous dense and small targets are present.
最近,分段任何模型3 (SAM 3)
Recently, Segment Anything Model 3 (SAM 3)
提出了一种统一分割和识别的方法
was proposed, unifying segmentation and recognition in a
promptable框架。
promptable framework.
在本文中,我们提出了一个初步的
In th