What is a dilated (atrous) convolution?
mediumAnswer
- Insert (d-1) zeros between kernel elements to enlarge the receptive field without adding parameters or reducing resolution.
- A 3×3 kernel with dilation 2 has an effective 5×5 view.
- Standard trick in segmentation networks (DeepLab, HRNet) where downsampling would lose spatial detail.
- Stacked dilated convs with growing dilation cover a large receptive field cheaply — also used in WaveNet for 1D audio.
Check yourself — multiple choice
- Dilation reduces receptive field
- Insert zeros between kernel elements → bigger receptive field, no extra params, no resolution loss
- Same as strided convolution
- Only for 1D signals
Dilated conv: gaps in kernel → bigger receptive field cheaply.
#cnn#segmentation
Practise Deep Learning
214 interview questions in this topic.
Related questions
- What is a transposed convolution ('deconv') and its checkerboard artifact issue?
- What inductive biases do CNNs have?
- What is the receptive field in a CNN?
- Compute the output shape of a 2D convolution with input (H, W), kernel k, stride s, padding p, dilation d.
- SAME vs VALID vs REFLECT padding — what are the practical differences?
- How does a depthwise separable convolution reduce compute?