Description
A Convolutional Neural Network (CNN) is a type of artificial neural networks that is primarily used for processing and analyzing visual data, such as images and video frames. CNNs have become widely popular in computer vision tasks due to their ability to automatically learn and extract meaningful features from input images. They consist of multiple layers, including convolutional layers, pooling layers, and fully connected layers, which work together to detect and recognize patterns in the data.
Today, Stiff Waist Beast wants to implement a small CNN, which processes an input image that is represented as a matrix of pixels with the following steps:
1. Horizontally flip the input image to enable the CNN to handle images with varying orientation.
2. Apply convolution using a filter to the flipped image.
The convolution is defined as follows:
Given an n * n image A and an m * m filter B, the resulting feature map after convolution is another (n - m + 1) * (n - m + 1) matrix C, where each element of C is computed as
Finally, Stiff Waist Beast is curious about the number of pixels in the resulting feature map whose values exceed a given threshold k (for fun only ^^).
As a poor student, there are still many tasks waiting for Stiff Waist Beast to handle in the association. He asks you to finish this task.
Input
The first line contains an integer T, representing that there will be T testcases.
The first line of each testcase has three integers n, m, k.
Each of the following n lines has n integers, aij, representing image A.
And each of the following m lines has m integers, bij, representing filter B.
1 <= T <= 10, 1 <= n <= 500, 1 <= m <= min(n, 5), 0 <= aij, k < 256, -5 <= bij <= 5.
Output
Print an integer: the number of elements in the resulting feature map C whose values exceed the given threshold k.
Don't forget to print '\n' in the end!