The Enhanced-Super Resolution Generative Adversarial Networks is an enhancement of Super-Resolution Generative Adversarial Networks by tweaking the model architecture to achieve high resolution. ISRGAN aims to further improve the quality of the image produced by the model by utilizing specially trained instances to upscale different portions of the image by enhancing each portion of the image by a model that is specially trained for such certain objects or classes. The idea is to divide and conquer the super-resolution problem utilizing the specialized models to up-scale sub-problems and improving the quality of generated images. Firstly image is passed through the Object detection phase which utilizes the Yolov3 structure to identify different classes present in the object, each class is then given to a generator that is specialized in that domain to further improve the quality. For the objects having unidentified classes or the base background image, we will have a generalized generator which will be trained on a combination of different domains. Also, to reduce the hardware requirement and improve the efficiency, we developed a way to split the images into sub-images to be enhanced individually and combined together to obtain the final image. These small images are in the form of squares which are enhanced and with the help of specialized generators and base models it is intended to convert low-resolution images into higher resolution models by up-scaling them to 4 times.

Super Resolution, Generative Adversarial Networks, Image enhancement, Upscaling

[1] X. Wang, K. Yu, S. Wu, J. Gu, Y. Liu, C. Dong, Y. Qiao, C.C. Loy, “ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks”, In the Proceedings of the 2018 European Conference on Computer Vision, Munich, Germany.
[2] C. Ledig, L. Theis, F. Husz´ar, J. Caballero, A. Cunningham, A. Acosta, A. Aitken, A. Tejani, J. Totz, Z. Wang, S. Wenzhe, “Photo-realistic single image super-resolution using a generative adversarial network”, In the Proceedings of the 2017 Conference on Computer Vision and Pattern Recognition.
[3] J. Redmon, A. Farhadi, “YOLOv3: An Incremental Improvement ”, In the Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition Workshops.
[4] J. Redmon, S. Divvala, R. Girshick, A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection”, In the Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops.
[5] C. Dong, C.C. Loy, K. He, X. Tang, “Learning a deep convolutional network for image super-resolution”, In the Proceedings of the 2017 European Conference on Computer Vision.
[6] B. Lim, S. Son, H. Kim, S. Nah, K.M. Lee, “Enhanced deep residual networks for single image super-resolution”, In the Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops.
[7] Y. Wang, F. Perazzi, B. McWilliams, A. Sorkine-Hornung, O. Sorkine-Hornung, C. Schroers, “A Fully Progressive Approach to Single-Image Super-Resolution”, In the Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition Workshops.
[8] V. Ferrari, M. Hebert, C. Sminchisescu, Y. Weiss, “Computer Vision” In the Proceedings of the 2018 European Conference on Computer Vision, Munich, Germany.
[9] N.S. Lele, "Image Classification Using Convolutional Neural Network", International Journal of Scientific Research in Computer Science and Engineering, Vol.6, Issue.3, pp.22-26, 2018