derek park
derek park
added below code(Pixel unshuffle code) before conv_first and use RealESRGAN_x2plus weight  if (OUT_SCALE == 2) { // Pixel unshuffle. int h = INPUT_H / OUT_SCALE; int w = INPUT_W...
It was officially made into an x4 model version. And later on request, how to use the x2 model version was explained in a previous comment https://github.com/wang-xinyu/tensorrtx/issues/1085#issuecomment-1229422132.