Robust Video Matting (RVM) Windows Installation Tutorial

My main channel where I introduce the latest fascinating AI tools

Related video for info/explanation

RVM - Robust Video Matting

I didn't write a GitHub page because video matting live would be overshadowing this soon.
Commands:
conda create -n RVMatting python=3.6
conda activate RVMatting

20 series or earlier
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.2 -c pytorch

30 series
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge

[suganesan]

What about the 40 series like rtx 4070 super gpu

[GOAT.]

I get AssertionError: Torch not compiled with CUDA enabled...

[nomberfax]

(RVMatting) R:\Useful Software\Work again\RWM>python inference.py --variant mobilenetv3 --checkpoint --device cuda --input-source input/da.mp4 --output-type video --output-composition output/a1.mp4 --output-alpha output/alp.mp4 --output-foreground output/for.mp4 --output-video-mbps 10 --seq-chunk 1
Traceback (most recent call last):
File "inference.py", line 18, in <module>
from torchvision import transforms
File "R:\Useful Software\Work again\Zadnyi\envs\RVMatting\lib\site-packages\torchvision\__init__.py", line 7, in <module>
from torchvision import datasets
File "R:\Useful Software\Work again\Zadnyi\envs\RVMatting\lib\site-packages\torchvision\datasets\__init__.py", line 1, in <module>
from .lsun import LSUN, LSUNClass
File "R:\Useful Software\Work again\Zadnyi\envs\RVMatting\lib\site-packages\torchvision\datasets\lsun.py", line 2, in <module>
from PIL import Image
File "R:\Useful Software\Work again\Zadnyi\envs\RVMatting\lib\site-packages\PIL\Image.py", line 114, in <module>
from . import _imaging as core
ImportError: DLL load failed: Не найден указанный модуль.

[wkfjvlzv]

Every time I use it
conda create -n RV Matting python=3.6
conda activate RV Matting
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forgepip install -r requirements_inference.txt
Is this something you should always do?

[ronlinetutorialssurvive]

ImportError: DLL load failed: Не найден указанный модуль.

[jmk]

Is it possible to batch run video files with this? Thanks

[jobennebur]

Great tutorial! But my cpu goes 100% while my 3060 is bearly used. I used the correct cudatoolkit installer

[jsserch]

Help me understand why would someone make something like this?. Who sits at their computer and makes a program that will mostly be used by "artists" and assumes they will have the knowledge, the skills and the time to do half the job?, download extra software, write command lines and compile a program. Why not do the rest of the job, create an intuitive user interface and sell it?.

[SitinprettyProductions]

Thanks for this in depth tutorial. I get an error message saying "modultnotfounderror: no module named torch" - any advice on what I should do?

[jakobtehwookiee]

Is there any way to use this while Streaming? or would that be too demanding on the GPU? It doesn't have to be 4k

[cg.man_aka_kevin]

Why there's just Anaconda Distribution version? No Individual Edition in Google?

[minibobber]

please make a tutorial for livestreaming/virtual webcam

[niu]

error: inference.py: error: the following arguments are required: --output-type

(RVMatting) C:\ANACONDA Codes\RVM> --output-type video --output-composition output/com.mp4 --output-alpha output/alp.mp4 --output-foreground output/for.mp4 --output-video-mbps 10 --seq-chunk 1
'--output-type' is not recognized as an internal or external command,
operable program or batch file.

what is happening?

[vodkaru]

Neat. Saved me collab hassle and wait.
Question though. That python inference command. The file contains all these commands but they're commented. Is there a way to uncomment them and just run the inference file. I tried remove the comments but got a syntax error
What does seq chunk number do? The github isn't exactly clear to me what it does? 1 is better or the higher the better?

Is this how I added downsample_ratio to the code line? --downsample-ratio 0.6

[bloomp]

ho i've done all and its says i have not NVidia.. i had hopes because it works live in the Browser !

[kunshail]

Is there any way to generate transparent background video with video matting?

[multivrsum]

Will it run on a NV-Quadro P5000 / vers:445.87 ?

[RioWebFest]

This is insanely good! Life saver. Thank you so much

[PeopleVersusTV]

I'm stuck on 'Solving environment'. There's a few lines of it failing to solve and then it just seems to get stuck. Anyone got any ideas?

Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: -

[Zanroff]

Is there a way to automatically do this to every video in a folder? Renaming the files is boring.