Top Highlights

• stride of 1 and no padding
• especially after the spatial dimensions are reduced by CNN layers, we can use another type of CNN layers that can increase (upsample) the spatial dimensions of intermediate feature maps. In this section, we will introduce transposed convolution, which is also called fractionally-strided convolution (Dumoulin and Visin, 2016), for reversing downsamp...
• In semantic segmentation that classifies at pixel-level, it will be convenient if the spatial dimensions of the input and output are the same. For example, the channel dimension at one output pixel can hold the classification results for the input pixel at the same spatial position.
• Each intermediate result is a ( � ℎ + � ℎ − 1 ) × ( � � + � � − 1 ) tensor that are initialized as zeros. To compute each intermediate tensor, each element in the input tensor is multiplied by the kernel so that the resulting
• In contrast to the regular convolution (in Section 7.2) that reduces input elements via the kernel, the transposed convolution broadcasts input elements via the kernel, thereby producing an output that is larger than the input

Domain