知识卡片
联合编码解码压缩上下文
编码解码器在低分辨率层聚合更大范围信息,再恢复到目标输出网格,可减少中间计算。瓶颈限制可传递信息并促使抽象,但也可能删去任务所需细节。压缩维度、表示精度与任务监督共同决定收益,不能把所有压缩都视为自动获得有用抽象。
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
One powerful thing you can do with encoders and decoders is to put them together, forming an encoder-decoder. Such an architecture first encodes the input image into a low-dimensional representation, then decodes the representation back into an image output. This is therefore a suitable architecture for image-to-image problems, and it has a few advantages over the image-to-image CNNs we saw in section 24.8: 1) by
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:摘录中关于联合编码解码压缩上下文的定义、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:One powerful thing you can do with encoders and decoders is to put them together, forming an encoder-decoder. Such an architecture first encodes the input image into a low-dimensional representation, then decodes the representation back into an image output. This is therefore a suitable architecture for image-to-image problems, and it has a few advantages over the image-to-image CNNs we saw in section 24.8: 1) by