"Exception: Unsupported source type"

[charpng]

import pyiqa
model = pyiqa.create_metric('unique', as_loss=False)
score = model(img)
#img = ..\image\10_1_C_1000_far_env1000.png

An error was encountered while running the above code

  File "D:\Tianma\UNIQUE\unique_.py", line 26, in <module>
    score = model(img)
            ^^^^^^^^^^
Exception: Unsupported source type

This is a png image with a resolution of 5218*2160. What do I need to change to solve this error?

[zwx8981]

Try: 1, If your image is a gray-scale one, convert it to RGB by duplicating it into three channels. 2, Resize your image such that the longest edge is lower than 1024.

[charpng]

Try: 1, If your image is a gray-scale one, convert it to RGB by duplicating it into three channels. 2, Resize your image such that the longest edge is lower than 1024.

Thank you very much, I successfully solved this problem. However, I'm facing a new problem now:
I have a set of images and corresponding MOS scores. How do I convert the data into a format for UNIQUE training? How would I get 'y', 'std1', 'std2', and 'yb' from the code mentioned in ImageDataset.py? I thought 'std' meant the standard deviation of the MOS score, but it seems not to be the case? Thank you.

sample = {'I1': I1, 'I2': I2, 'y': y[0], 'std1':y[1], 'std2':y[2], 'yb': y[3]}

[zwx8981]

std is the standard deviation of the MOS. You may refer to one of the scripts in IQA_Database (e.g., data_koniq10k.m) to generate these data.