Computationally Efficient Approaches for Image Style Transfer

Pandey, R. K.; Karmakar, Samarjit; Ramakrishnan, A. G.

doi:10.48550/arxiv.1807.05927

Cited by 2 publications

(1 citation statement)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These approaches solve an optimization problem and hence, are slower. Justin Johnson et al [15] and Pandey et al [16] use the benefits of perpixel as well as perceptual loss funtions and propose a computationally efficient, optimization-free approach that provides results for image transformation tasks that are qualitatively similar to those of the above optimization-based approaches. The super-resolution algorithm SRGAN [8] uses a weighted combination of three different loss functions, namely mean square error, perceptual and adversarial loss to obtain a sharper reconstruction.…”

Section: Related Workmentioning

confidence: 99%

MSCE: An Edge-Preserving Robust Loss Function for Improving Super-Resolution Algorithms

Pandey

Saha

Karmakar

et al. 2018

Neural Information Processing

Self Cite

View full text Add to dashboard Cite

With the recent advancement in the deep learning technologies such as CNNs and GANs, there is significant improvement in the quality of the images reconstructed by deep learning based superresolution (SR) techniques. In this work, we propose a robust loss function based on the preservation of edges obtained by the Canny operator. This loss function, when combined with the existing loss function such as mean square error (MSE), gives better SR reconstruction measured in terms of PSNR and SSIM. Our proposed loss function guarantees improved performance on any existing algorithm using MSE loss function, without any increase in the computational complexity during testing.

show abstract

Section: Related Workmentioning

confidence: 99%