Some Studies for Rotational Invariance in Convolutional Neural Networks.

Date of Submission

December 2018

Date of Award

Winter 12-12-2019

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science


Electronics and Communication Sciences Unit (ECSU-Kolkata)


Das, Swagatam (ECSU-Kolkata; ISI)

Abstract (Summary of the Work)

Computer vision is an amazing field of using computing machinery to resemble human vision system where common tasks include image recognition. Convolutional Neural Networks(CNNs) has shown amazing capabilities for such tasks. Besides several advantages, CNNs are significantly susceptible to the rotational transformation of images. CNNs from the input images tries to abstract each image by learning local knowledge using convolutional filters applied over all parts of the image. We propose two methods for considering rotational invariance in learning. The first method uses rotational invariant filter method. The second method uses nonlinear neighbourhood component analysis for learning specialized filters for which the two images of the same class but in different rotational space lie close to each other. We discuss future scope solution for the NCA method.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


This document is currently not available here.