VMFormer
VMFormer copied to clipboard
SHI-Labs
[Preprint] VMFormer: End-to-End Video Matting with Transformer
Hello, I encountered a problem when running the inference script. I tried to apply the matting to my own image, but the output in the console shows a warning like...
The links for the 512x288 and 1920x1080 model weights are the same. From experimentation, it seems like the both link to the 512x288 weights.
follow "https://github.com/SHI-Labs/VMFormer/blob/main/INSTALL.md" Step 5: Install CUDA version of MultiScaleDeformableAttention cd ./models/ops sh ./make.sh python test.py test.py will error “ "error in ms_deformable_col2im_cuda: too many resources requested for launch\n",” 
``` def build_vm(args): device = torch.device(args.device) if 'swin' in args.backbone: from .swin_transformer import build_swin_backbone backbone = build_swin_backbone(args) else: backbone = build_backbone(args) ``` swin_transformer isn't a defined requirements. I've scoured the...
gcc -std=gnu99: No such file or directory nvcc fatal : Failed to preprocess host compiler properties. error: command '/usr/local/cuda/bin/nvcc' failed with exit status 1
Hi! I wonder if the to-do list is being worked on or if this project is abandoned? Best Regards Ludvig Friberg
Hello, when I was running inference_vm.py, this error would occur whenever num_frames was set larger than 1. I wonder how can I solve this problem, thank you！
