GoogLeNet: Counting Convolution Parameters
Beginner
The problem, from first principles
GoogLeNet (Szegedy et al., 2015) builds inception modules from several parallel convolutions, using 1x1 convolutions to cut channels before the bigger kernels. Counting parameters per convolution shows where the cost sits.
From theory to code
Implement conv_params(c_in, c_out, k), the number of weights and biases in one convolution layer.
Constraints
- Returns an integer.
Hints
Hint 1
Multiply input channels, output channels and the kernel area, then add one bias per output channel.
Keywords
research-paperscomputer-visioncnninception
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.