📌 📄 内容摘要:
liuweiqing (@14790897) 在 今天第一次触发gemini3pro复读 中发帖
[image]
[image]
飞思 Okay, SAM 3 (Segment Anything Model 3) by Meta is out (released late 2025 according to the context). It introduces "promptable concept segmentation" (open-vocabulary segmentation using text, like "yellow school bus"). It unifies image and video segmentation. It seems to use a "shared perception encoder (PE) vision backbone", and a "detector and tracker". The detector part....
────────── 链接信息 ──────────
🔗 论坛链接: linux.do
📎 访问地址: https://linux.do/t/topic/1683369/1
─────────────────────────────
📢 来源：LINUX DO
