Single Image Super-Resolution Using GANs | Lecture 68 (Part 2) | Applied Deep Learning

preview_player
Показать описание
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

Рекомендации по теме
Комментарии
Автор

Hellor Sir, i have a question about the balance between these terms loss:
(w1*MSE + w2*Perceptual_Loss + w3*Gan_Loss),
how do we set these balance weight for each of term, thank you sir.

DangNguyen-yhmm
Автор

00:07:43 sir when we say kernal size 9 is it 3×3 or 9×1 or 1×9 ? Plz excuse this dumb question

missingapps