
Answer-first summary for fast verification
Answer: Feature Cross
Feature Cross involves creating new features by combining (crossing) two or more existing features, which introduces non-linearity to the model. However, in the context of signature verification, this technique is unnecessary and does not contribute to the accuracy of distinguishing between genuine and forged signatures. Kernel Selection refers to the computation on a sub-matrix of pixels, Stride is determined by sliding the kernel by 1 pixel, and a Max pooling layer simplifies the model by taking the maximum value of a small region. These are all applicable and beneficial in CNNs for image recognition tasks like signature verification.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
In a project aimed at enhancing bank security, your team is tasked with developing a system to distinguish between genuine and forged signatures on checks and documents by comparing them against signatures stored in the bank's database. Given the rarity of forged signatures, the system must not only accurately identify the customer but also determine with high precision whether the signature is genuine or a skilled forgery. The solution will utilize a Convolutional Neural Network (CNN) for image recognition. Considering the need for high accuracy and the specific challenges of signature verification, which of the following technical specifications is unsuitable for use with a CNN in this context? Choose one option.
A
Kernel Selection
B
Max pooling layer
C
Stride
D
Feature Cross
No comments yet.