Feb.17 Informal Response
Victoria Yuanyuan Chang
A. Convolutions
- filter = [ [-1, -2, -1], [0, 0, 0], [1, 2, 1]] This filter accentuates vertical lines in the image.
Output:
- filter = [ [-1, 0, 1], [-2, 0, 2], [-1, 0, 1]] This filter accentuates horizontal lines in the image.
Output:
- filter = [ [1, 0, 0], [ 0, 0, 0], [0, 0, -1] ] This filter darkens the image.
Output:
4.The application of convolutions is helpful for computer vision is that it can extract features of the image so it’s easier to label them according to these features.
B. Pooling
The following is the image with the filter that accentuate vertical line after pooling:
Pooling reduces irrelevant information in the image so that the relevant features can be further emphasized. This pooling filter takes iterates over all the pixels, taking every other one, and find the largest pixel among itself and its three neighbors. It creates a new array that is half the size of the input image, and extract that largest pixel into the output array. The method results in reduced size.