Closed BracerJack closed 3 years ago
4K I would use --model-backbone-scale 0.125
. Sample pixels 320000 is fine.
Can I confirm that the whole improved statement would now be: python inference_video.py --model-type mattingrefine --model-backbone resnet50 --model-checkpoint ./model/PyTorch/pytorch_resnet50.pth --model-backbone-scale 0.125 320_000 --video-src ./input/Raw.avi --video-bgr ./input/Empty_Background.png --output-dir ./output --output-types com fgr pha err ref
Thanks.
You missed --model-refine-sample-pixels
. Other than that it's good.
Where do I put that ? Before or after the --model-backbone-scale 0.125 320_000 ? Do I put the 0.125 320_000 before the --model-backbone-scale but after the --model-refine-sample-pixels or do I put it after ? Or do I somehow combine the two to form the statement ?
Can you give me the full complete statement so I don't have to guess ? Thanks :D
This should be good, but I haven't tested it myself.
python inference_video.py --model-type mattingrefine --model-backbone resnet50 --model-checkpoint ./model/PyTorch/pytorch_resnet50.pth --model-backbone-scale 0.125 --model-refine-sample-pixels 320_000 --video-src ./input/Raw.avi --video-bgr ./input/Empty_Background.png --output-dir ./output --output-types com fgr pha err ref
I am currently using: python inference_video.py --model-type mattingrefine --model-backbone resnet50 --model-checkpoint ./model/PyTorch/pytorch_resnet50.pth --model-refine-sample-pixels 320_000 --video-src ./input/Raw.avi --video-bgr ./input/Empty_Background.png --output-dir ./output --output-types com fgr pha err ref
So far it works VERY WELL.
But sometimes when the gap is small, the cut out will not be made probably because the sample is too low ? What modifications should I make to the command line statement that I use on top to make it really good for 4K footage ? I don't care if it is not real time, I care more about the cut out quality, it can take AS LONG AS IT TAKES, Quality comes first :D