Rahul Deora

I am guessing it will fail in most cases as the datasets they trained is very limited and mostly humans. But author may have better response after his/her testing

@swelcker How it the quality of the output? How does it compare to Google Meet's background removal which requires no background image?

Can you share your data @xiongzhu666 ? Would love to work on sky segmentation

@xiongzhu666 Which sky matting data are you using?

Hi @dennisbappert great work! Can you explain how channels_last memory format exactly helps? I had tried with some augmentations from albumentations earlier however the results degraded and predictions become un-confident....

Thanks! Will have to look into this abit. You mentioned 'e.g. perspective warping, greyscale, blur, jitter' augmentations but did not show any of those above. Did you end up applying...

Thanks, will go through this :)

Hi Jack, the link you shared does not seem to be loading. Can you recheck it?

The domain http://cv-datasets.oppoer.me also does not work

@JackSyu Any update on the data?