Marco Forte

Hi, can you expand on the problem you're having? Do you see any output image at all? Right now it's programmed to output the mask overlaying the original image.

Hi, I don't have the results on hand right now. In the paper we report very good results around epoch 20, so I suspect the results are decent around epoch...

Hi everyone, I'm very sorry to say I won't be able to release the training code anytime soon. I hope you can find the pretrained model useful. Many recent works...

I recommend these works which have their source code available - https://grail.cs.washington.edu/projects/background-matting-v2/ - https://github.com/Yaoyi-Li/GCA-Matting - https://github.com/JizhiziLi/animal-matting

Good question. When compositing a foreground above one another we essentially solve for the foregrounds three times. 1. Solve for each foreground independently. 2. Resize/crop the second foreground to be...

Instead of only feeding the network the binary trimap we also feed the distance transformed version. The distance from the definite foreground and background regions is a strong indicator of...

Hi I used the ResNet-50 weights from this repo https://github.com/joe-siyuan-qiao/pytorch-classification/tree/e6355f829e85ac05a71b8889f4fff77b9ab95d0b The finetuning we refer to is just dropping the learning rate and training for more epochs.

Hi, we use their ResNet-50 weights from pre-training on ImageNet, http://cs.jhu.edu/~syqiao/WeightStandardization/R-50-GN-WS.pth.tar For dropping the learning rate here is the relevant text in the paper, "_The initial learning rate is set...

It is trained on inputs of size 320x320 but these also include larger size patches which have been downsampled to 320x320. The model performs well on images of 2000x2000. You...

Hi thanks for your interest and taking time to inform me of this issue. As raphychek pointed out it has been corrected already, see https://github.com/MarcoForte/FBA_Matting/issues/7