43 lines
1.1 KiB
Python
43 lines
1.1 KiB
Python
import folder_paths
|
|
import torch
|
|
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry
|
|
from einops import rearrange, repeat
|
|
import requests
|
|
from PIL import Image
|
|
import io
|
|
import numpy as np
|
|
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry, SamPredictor
|
|
from einops import rearrange, repeat
|
|
|
|
class SAM_Remote_Emb:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {
|
|
"required": {
|
|
"model_type": (["vit_h", "vit_l", "vit_b"],),
|
|
"ckpt": (folder_paths.get_filename_list("sam"),),
|
|
},
|
|
}
|
|
|
|
RETURN_TYPES = ("SAM", "SAMPREDICTOR")
|
|
RETURN_NAMES = ("sam", "predictor")
|
|
|
|
FUNCTION = "segment"
|
|
|
|
CATEGORY = "image"
|
|
|
|
def segment(self, model_type, ckpt):
|
|
ckpt = folder_paths.get_full_path("sam", ckpt)
|
|
sam = sam_model_registry[model_type](checkpoint=ckpt)
|
|
predictor = SamPredictor(sam)
|
|
|
|
return (sam, predictor)
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
|
"SAM_Loader": SAM_Remote_Emb
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"SAM_Loader": "SAM Loader"
|
|
}
|