Inpaint error - You are trying to inpaint without masks input

[SQLCODE917]

• Running locally on Win11/4090 w/python 3.10.5 on pyenv

• Added video 10 second 720p video, selected 3 masks, clicked Tracking, success:

• Clicked inpaint, error like Error! You are trying to inpaint without masks input. Please track the selected mask first, and then press inpaint. If VRAM exceeded, please use the resize ratio to scaling down the image size.

  • Tried restarting from the beginning and setting the resizing ratio to .5, same error.

Is this the correct order of operations? Is there a step I'm missing to send the same masks that just worked for tracking to inpaint?

[potatoker]

same error here. with no useful log to deal with

[memoryunreal]

Please try to set the resize ratio slider to 0.1. If it works, it means the GPU memory is exceeded in the previous operation.

[potatoker]

Finally get through this by doing the following:

1. cut my whole uploaded video to shorter duration, resize ratio to smaller, click getInfo
2. add the mask step by : click the object in the left "Image" box; click "Add mask" button above; click "Tracking"; click "inPainting".（I previously misunderstood the "add mask" button to click first before select object)

Not sure If it is the vram memory problem or the step problem

[zhanghongyong123456]

Finally get through this by doing the following:

1. cut my whole uploaded video to shorter duration, resize ratio to smaller, click getInfo
2. add the mask step by : click the object in the left "Image" box; click "Add mask" button above; click "Tracking"; click "inPainting".（I previously misunderstood the "add mask" button to click first before select object)

Not sure If it is the vram memory problem or the step problem

It was vram problem in my test just now. When I set 0.1, I could run inpaint. It worked, but the effect was poor

[memoryunreal]

We will add a VRAM monitor in the future, and sorry for the lack of an operation tutorial (incoming). For the VRAM problem, we are trying to optimize the RAM usage and VRAM usage.

[zhanghongyong123456]

We will add a VRAM monitor in the future, and sorry for the lack of an operation tutorial (incoming). For the VRAM problem, we are trying to optimize the RAM usage and VRAM usage.

https://github.com/gaomingqi/Track-Anything/issues/21 My problem is still there

[SQLCODE917]

Please try to set the resize ratio slider to 0.1. If it works, it means the GPU memory is exceeded in the previous operation.

Unfortunately, the error persists, even at 0.1 ratio

I have also tried to cut the source video to 5 seconds, making it shorter than the 8 second parkour inpainting example video, set scale to 0.1, and it's still producing the same error

[zhanghongyong123456]

Please try to set the resize ratio slider to 0.1. If it works, it means the GPU memory is exceeded in the previous operation.

Unfortunately, the error persists, even at 0.1 ratio

I have also tried to cut the source video to 5 seconds, making it shorter than the 8 second parkour inpainting example video, set scale to 0.1, and it's still producing the same error

Maybe yours video's resolution two high, i test video 640X360,set Raise ratio=1, raise same error,Until the Raise ratio is set to 0.5,can run,you can set raise ratio 0.05 or lower

[Beckjiang]

I tried it and works with ratio 0.5, but the masked objects are still in the output video.

[finnschi]

Same issue RTX4090 (24Gb Vram) , tried to inpaint a 480x270p video thats only 25 f long and nothing, same error as everyone seems to be getting

I am on windows

[SQLCODE917]

I pulled latest master today, scaled the video down to 360x180 and tried resize ratio all the way down to 1, 0.5 and even 0.02. I tried using only 1 mask - no luck, same error.

I also tried the lowest scale with the sample videos, just to be sure it's not just my own video - none of the sample videos I was able to inpaint.

[gaomingqi]

@SQLCODE917 Hello. In the current version, inpainting cannot work without tracking results. So if you want to 'remove' an object in a video, first you should select and add the object mask, then click 'tracking' to locate the selected object throughout the video, and finally, click 'inpaint' for the inpainting results. Thanks.

[SQLCODE917]

Thank you, yes - those are the steps that I have been using

[WelschCode]

It gives me an error if I try to mask then inpaint. It just says "Error" and I crashes.

[memoryunreal]

It gives me an error if I try to mask then inpaint. It just says "Error" and I crashes.

After you mask a target, you should track it first to get masks in the following frames and then inpaint it in the whole video