How to run inference on single images?

[ericzzfu]

Also trying to figure it out...

[halqadasi]

Is there any answer to this thread? I am also stuck to find the inference for a single image.

I spent long time understanding the code, finally I realized that I should provide the framework with both HR and LR to evaluate it. There is no option to test it without a ground-truth dataset.

[Francis0625]

We will update the code for the inference on a single image as soon as possilble.

[yejing0]

I also have same question

[anewusername77]

any update?

[ForeUP]

I write a simple code to test on a single image, just like below: Replace the code in https://github.com/Francis0625/Omni-SR/blob/6f5e53b2e0f0afb20901b4270c0a7a08e0a54d1d/test_scripts/tester_Matlab.py#L135 by:

test_single = True
        with torch.no_grad():
            if test_single:
                from PIL import Image
                from torchvision import transforms as T
                save_dir = "test/result/dir"
                transform_valid = T.Compose([
                    T.ToTensor(),
                    T.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)) ])
                img_path = "image/to/test.png"
                img = Image.open(img_path)
                img_ = transform_valid(img)
                image = img_.unsqueeze(0)
                image = image.cuda()
                res = self.network(image)
                res = tensor2img(res.cpu())
                temp_img = res[0, :, :, :]
                temp_img = cv2.cvtColor(temp_img, cv2.COLOR_RGB2BGR)
                name = os.path.basename(img_path)
                cv2.imwrite(os.path.join(save_dir, 'test_in_author_{}'.format(name)), temp_img)
                final_psnr = 0.0
                final_ssim = 0.0
            else: