ResNet: The Convolution Output Size
Beginner
The problem, from first principles
ResNet (He et al., 2016) opens with a large strided convolution that shrinks the image. Knowing how the spatial size changes at each layer is needed to design such networks and to check their shapes.
From theory to code
Implement conv_output_size(n, k, s, p), returning the output length of a convolution along one dimension.
Constraints
- Use integer division.
Hints
Hint 1
Add the padding on both sides, subtract the kernel, divide by the stride, floor, then add one.
Keywords
research-paperscomputer-visioncnnarchitecture
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.