Question 9In Depthwise Separable Convolution you:

In Depthwise Separable Convolution, you perform two steps:

1. Depthwise Convolution: This is a process where an input channel is convolved with its own set of filters, as opposed to the standard convolution where each input channel is convolved with all filters. This helps in reducing the computation and also captures spatial features better.

2. Pointwise Convolution: After the depthwise convolution, a pointwise convolution, which is a normal convolution with a 1x1 filter, is applied. This is used to build new features through computing linear combinations of the output of the depthwise convolution.

These two steps together form the Depthwise Separable Convolution. It is mainly used in mobile applications due to its efficiency.

In Depthwise Separable Convolution, you perform two steps:

1. Depthwise Convolution: This is the first step where you apply a single filter per input channel. This means that each channel in the input data gets its own single convolutional filter. This step helps in capturing the spatial features of the input data.

2. Pointwise Convolution: This is the second step where you apply a 1x1 convolution to the output of the depthwise convolution. This step is used to build new features through computing linear combinations of the input channels.

These two steps together form the Depthwise Separable Convolution. This method is computationally efficient and also reduces the model size without sacrificing the model's performance. It is widely used in mobile applications where computational resources are limited.