the refined mask is even worse than SAM

[buptlj]

input image:

input mask(SAM result):

SAM result:

segrefiner result:

[EricShenYuQi]

单个图片的结果不能说明什么问题吧，还是要看统计指标上的结果感觉

[buptlj]

单个图片的结果不能说明什么问题吧，还是要看统计指标上的结果感觉

我在自建的测试集上测了，以SAM的结果为输入，segrefiner输出结果并没有明显改善，部分边界确实更好了，但会多分割出来其他区域，总体主观评测下来效果不佳

[gufran-barcha]

input image:

input mask(SAM result):

SAM result:

segrefiner result:

Hello, Can you tell me how to do inference for single image or how you have acheived this. As there are no instruction on it. Also do we need to train a dataset for inference or just we can do inference without it . I was just checking the repo and I didn't really understand . How to do inference for single image.

Will appreciate your help.