PIN IMAGE DOWNSAMPLE
Beginner
Difficulty: Easy Tags: Computer Vision
Story
Pinterest downsamples every pin image before it reaches the visual-search backbone. This max-pool layer is the very first operation in that backbone, run on millions of images a day.
The Math
2D max pooling with kernel size k, stride s, no padding: output size
floor((H - k) / s) + 1 per dimension, each output cell is the max over its receptive field.
Input Format
H W k s
p_1,1 ... p_1,W
...
p_H,1 ... p_H,W
Output Format
Pooled matrix, row by row, matching the input's own numeric type (integer values stay integer, otherwise 6-decimal floats).
Constraints
1 <= H, W <= 512,1 <= k <= min(H, W),1 <= s <= k- Time limit: 1.0 second.
Example
Input
4 4 2 2
1 3 2 4
5 6 7 8
9 1 2 3
4 5 6 0
Output
6 8
9 6
Keywords
computer-vision
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.