This project is part of course CS F425 Deep Learning at BITS Goa.
In this project, we learned that Cycle Consistency Adversarial Networks tries to solve image-image translation problems without paired data. The objective function it uses is the Adversarial losses and Cycle consistency losses. Image-Image translation is used in Object Transfiguration, Photo Enhancement, Style Transformation, Season Transformation and Image Segmentation. We made the model and tested it on a new city dataset and performed image segmentation.
Addtional details about the project can be found in project report.
- Anurag Nagpal [2018B1A70939G]
- Sahdev Nuwal [2018B3A70900G]
- Varun Bankar [2018B2A70295G]
You can find Colab notebook related to this project here.
This project is based on the Paper titled “Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks” by Jun-Yan Zhu et al 2017.