SAM-Adapter-PyTorch
SAM-Adapter-PyTorch copied to clipboard
support for prompts
Hi, Great work. Does your code support prompts, like point-prompts or box prompts which are in the original SAM
Greetings! This project primarily addresses traditional image segmentation scenarios, such as shadow detection without prompt input. As a result, we have not integrated a prompt encoder into the neural network. However, you have the option to incorporate the prompt encoder from SAM if you wish to exercise control over the segmentation process, while still utilizing our modified SAM image encoder and decoder. Feel free to enable the prompt encoder and leverage its capabilities alongside our adapted SAM components.
Greetings! This project primarily addresses traditional image segmentation scenarios, such as shadow detection without prompt input. As a result, we have not integrated a prompt encoder into the neural network. However, you have the option to incorporate the prompt encoder from SAM if you wish to exercise control over the segmentation process, while still utilizing our modified SAM image encoder and decoder. Feel free to enable the prompt encoder and leverage its capabilities alongside our adapted SAM components.
想问一下如果我想手动添加提示给模型的话,会不会损坏sam2 adaptor本身的性能?我按照原版sam2的方式加入了添加点作为提示的部分,但是效果一直不太好,所以才来问问你们~希望可以得到回复!
I'd like to ask if adding manual prompts to the model could harm the SAM-2 adaptor's inherent performance. I've added functionality for point prompts following the original SAM-2 approach, but it hasn't been performing well. Could you shed some light on this? I'd appreciate your response!