Closed Satyaexp closed 10 months ago
Try using --jit --mode base-nightly
options. This will not solve the problem 100% but generates better results. Please wait for the first 30 seconds for model conversion for the first execution of --jit
option.
thanks will try
I tested it, as mentioned its not 100 percent but better then the earlier one. do you have any suggestion to make it perfect. also is there any issue with the images and the way they are shot. Any help to resolve this is highly appreciated. Willing to also engage professionally....
Sure thing. First, you need more simple background. Our model could struggle generating good quality masks when there are other objects in the background. Second, since your concern is to generate human mask, try removing objects like baby trolley. Let kids stand up by themselves. Lastly, try finetuning the model with human annotations only. I recommend making dataset with multiple people in the single image frame. You can train your model from our original repository InSPyReNet
Closing due to inactivity. Please open another issue if problem persists.
i have a few images where humans are getting cut, the accuracy of the other images that i have tested are perfect, i tested over 600000 images and out of which 10 images are creating an issue. can you help me fix this. i really want to continue using your model please help me sort this issue... attached images with issues
https://github.com/plemeri/transparent-background/files/13944242/issue.zip