通过预训练权重推理图片进行超分

[youngbaldy]

我已参考 https://huggingface.co/spaces/KyanChen/FunSR/blob/main/App_main.py 中的sr_func函数，打算对其他遥感数据集进行超分，但是得到的结果似乎是字典形式：{'encoder_spec': {'name': 'rdn', 'args': {'no_upsampling': True}}, 'has_multiscale': True, 'neck': {'name': 'transformer_neck', 'args': {'d_dim': 256, 'downsample': True, 'has_pe': True, 'has_norm': True, 'class_token': True, 'num_encoder_layers': 3}}, 'decoder': {'name': 'sirens', 'args': {'num_inner_layers': 5, 'is_residual': True}}, 'global_decoder': {'name': 'sirens', 'is_proj': True, 'args': {'num_inner_layers': 5, 'is_residual': True}}, 'encoder_rgb': True, 'n_forward_times': 1, 'encode_hr_coord': True, 'has_bn': True, 'encode_scale_ratio': True, 'local_unfold': False, 'weight_gen_func': 'bicubic'}，而我期望得到超分后的图像，我该怎么做？应该通过tools文件夹下的某个可视化脚本吗？

[youngbaldy]

上述问题已解决，模型输入似乎只能是4848，4倍超分输出是192192。这输入似乎是固定的？

[xiaying-bot]

你好，请问有提供模型的权重参数吗？我打开huggingface链接解压出来里面只有一个data.pkl，似乎没有UC_FunSR_RDN.pth文件？

[KyanChen]

https://huggingface.co/spaces/KyanChen/FunSR/tree/main/pretrain

[xiaying-bot]

https://huggingface.co/spaces/KyanChen/FunSR/tree/main/pretrain

非常感谢您的回复，我在前面README里的【5. [optional] Model Download】看到了这个链接，下载之后解压出来里面的文件结构是这样的，没有UC_FunSR_RDN.pth权重文件

[KyanChen]

[Bulinken]

你好，ulits模块中没有找到对rgb2ycbcr_pt函数的定义 def calculate_ssim_pt(img, img2, crop_border, test_y_channel=False, **kwargs): """Calculate SSIM (structural similarity) (PyTorch version). Paper: Image quality assessment: From error visibility to structural similarity The results are the same as that of the official released MATLAB code in https://ece.uwaterloo.ca/~z70wang/research/ssim/. For three-channel images, SSIM is calculated for each channel and then averaged. Args: img (Tensor): Images with range [0, 1], shape (n, 3/1, h, w). img2 (Tensor): Images with range [0, 1], shape (n, 3/1, h, w). crop_border (int): Cropped pixels in each edge of an image. These pixels are not involved in the calculation. test_y_channel (bool): Test on Y channel of YCbCr. Default: False. Returns: float: SSIM result. """

assert img.shape == img2.shape, f'Image shapes are different: {img.shape}, {img2.shape}.'

if crop_border != 0:
    img = img[:, :, crop_border:-crop_border, crop_border:-crop_border]
    img2 = img2[:, :, crop_border:-crop_border, crop_border:-crop_border]

if test_y_channel:
    img = rgb2ycbcr_pt(img, y_only=True)
    img2 = rgb2ycbcr_pt(img2, y_only=True)

img = img.to(torch.float64)
img2 = img2.to(torch.float64)

ssim = _ssim_pth(img * 255., img2 * 255.)
return ssim

[YHY0202]

我已参考 https://huggingface.co/spaces/KyanChen/FunSR/blob/main/App_main.py 中的sr_func函数，打算对其他遥感数据集进行超分，但是得到的结果似乎是字典形式：{'encoder_spec': {'name': 'rdn', 'args': {'no_upsampling': True}}, 'has_multiscale': True, 'neck': {'name': 'transformer_neck', 'args': {'d_dim': 256, 'downsample': True, 'has_pe': True, 'has_norm': True, 'class_token': True, 'num_encoder_layers': 3}}, 'decoder': {'name': 'sirens', 'args': {'num_inner_layers': 5, 'is_residual': True}}, 'global_decoder': {'name': 'sirens', 'is_proj': True, 'args': {'num_inner_layers': 5, 'is_residual': True}}, 'encoder_rgb': True, 'n_forward_times': 1, 'encode_hr_coord': True, 'has_bn': True, 'encode_scale_ratio': True, 'local_unfold': False, 'weight_gen_func': 'bicubic'}，而我期望得到超分后的图像，我该怎么做？应该通过tools文件夹下的某个可视化脚 请问你这个怎么解决的