Jindong Zhang

I think [the reply from the author](https://github.com/ckkelvinchan/RealBasicVSR/issues/22#issuecomment-1072090473) will help you! @automatic0313

PI是直接由NIQE和NRQM(Ma)计算得到(参考https://openaccess.thecvf.com/content_ECCVW_2018/papers/11133/Blau_2018_PIRM_Challenge_on_Perceptual_Image_Super-resolution_ECCVW_2018_paper.pdf )，NIQE和Ma是不需要gt的，细节可以看这里的代码https://github.com/roimehrez/PIRM2018/blob/master/utils/calc_scores.m#L33-L34 。 只不过https://github.com/roimehrez/PIRM2018 中这里计算的函数入口需要提供gt路径，这个gt路径只是用来计算MSE的。 简要来说，把videoLQ预测的图像放在'your_results'中，将[PIRM2018](https://github.com/roimehrez/PIRM2018 )的evaluate_results.m 和 utils/calc_scores.m与gt相关的代码都注释掉就可以了

mmediting支持了使用python计算NIQE, 但我的经验是在GPU上测试VideoLQ的NIQE也要2小时, 不知道你测试的整个数据集是否可行. 可以看看[IQA-PyTorch](https://github.com/chaofengc/IQA-PyTorch), 使用PyTorch支持了很多指标的GPU加速,比matlab要快.

具体问题可以到[这里](https://github.com/ckkelvinchan/RealBasicVSR)咨询作者, 在MMEditing中NIQE的测试结果是3.76~3.8

Closing due to inactivity, please reopen if there are any further problems.

Sorry for the late reply. I have succeeded to run `matting_tutorial.ipynb` in the environment `torch 1.8.0+cu101, mmcv(mmcv-full) 1.6.2, MMEditing 0.15.2`. Can you supply more information?

It looks like some bugs in colab? Because I can run `pred_alpha = matting_inference(model, merged_path, trimap_path) * 255` successfully on my PC. Returns of `data = collate([data], samples_per_gpu=1) ` are:...

We will not consider supporting it because it's lower priority. We can give you some help to add it to projects if you are willing.

Closing due to inactivity, please reopen if there are any further problems.

Please sign the CLA(Contributor License Agreement) @xuan07472