SAM 2:把分割推进到视频
SAM 2: Segment Anything in Images and Videos
Ravi et al. · 2024 · arXiv:2408.00714
Foundation paper · permanent
2024
图像上的可提示分割,能否在视频里保持目标身份不漂移?
What they did流式记忆(memory bank)+ 掩膜传播,用前一帧结果约束当前帧。
Evidence视频分割基准上优于此前方法;交互速度满足近实时。
Do not over-trust巡检视频里目标会被遮挡、出画再入画;记忆机制对“同一缺陷再次出现”并不可靠。
Limits长视频显存与算力开销大;遮挡后的重识别仍会失败;不提供缺陷类别。
In-class task取一段 2 分钟的巡检视频,统计掩膜漂移帧占比,并给出可接受的丢帧上限。
A foundation paper is the landmark work for a method and stays permanently: it explains why the method works, and is never replaced for being old.
© SailiTech · Printed at: