知识卡片
优网跳连传递高分辨率细节
U-Net把编码器同尺度的特征沿跳连送到解码器,并通常按通道拼接,使预测能利用细位置与粗上下文。拼接需要对齐空间尺寸,也增加通道和激活存储。跨越瓶颈的路径意味着整网不再仅依赖最小表示,不能声称仍保持严格单一低维信息瓶颈。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 卷积神经网络
原书出处 · 1 处
- · 查看原书依据
Encoder-decoders force the signal to pass through a bottleneck, and although this can be a good thing (as discussed above), it also makes the task of the decoder rather difficult. In particular, the decoder may fail to be able to output high frequency details; for a semantic segmentation network, the consequence could be that the predicted label map is very coarse.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第24章《卷积神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c024b.xhtml)
- 结论依据:摘录中关于优网跳连传递高分辨率细节的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Encoder-decoders force the signal to pass through a bottleneck, and although this can be a good thing (as discussed above), it also makes the task of the decoder rather difficult. In particular, the decoder may fail to be able to output high frequency details; for a semantic segmentation network, the consequence could be that the predicted label map is very coarse.