Ganesha Srinivas Damaraju
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Ganesha Srinivas Damaraju has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Ashly Ajith
Amrita Vishwa Vidyapeetham, IN
1 shared publication - Gopakumar Gopalakrishna Pillai
Department of Computer Science and Engineering, Amrita School of Computing, Amrita Vishwa Vidyapeetham, Amritapuri, India
1 shared publication
Latest Publications
- Open Access
- 15 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Improved Discrepancy based Domain Adaptation Network using combined loss functions and Feature transformations✱
Domain shifts are a common problem in computer vision. As a result, a classifier trained on a source domain cannot perform well on a target domain. Due to this, a source classifier trained to differentiate based on a specific distribution cannot classify well on new data from a different distribution. In this work, we have studied different domain adaptation networks for their complexity and performance and have shown that a simple discrepancy based network can be modified to produce a performance on par with the more complex GAN based networks[8]. This is achieved by picking and selectively combining loss functions and preprocessing transformations. We have shown that the proposed method has produced accuracies 77.62% and 81.67% on popular datasets Office-31 and MNIST/MNIST-M, surpassing the highest reported accuracies by 6.47% and 16.67% compared to seminal papers [16][1] in this field. 1
- 15 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessImproved Discrepancy based Domain Adaptation Network using combined loss functions and Feature transformations✱
Domain shifts are a common problem in computer vision. As a result, a classifier trained on a source domain cannot perform well on a target domain. Due to this, a source classifier trained to differentiate based on a specific distribution cannot classify well on new data from a different distribution. In this work, we have studied different domain adaptation networks for their complexity and performance and have shown that a simple discrepancy based network can be modified to produce a performance on par with the more complex GAN based networks[8]. This is achieved by picking and selectively combining loss functions and preprocessing transformations. We have shown that the proposed method has produced accuracies 77.62% and 81.67% on popular datasets Office-31 and MNIST/MNIST-M, surpassing the highest reported accuracies by 6.47% and 16.67% compared to seminal papers [16][1] in this field. 1