init project

checkpoints/ObjectAwareModel.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d99244a9c7a4d7330af28d1801ebee362fa469ac564b57c8eab48b43e4e64a1
+size 140395126

checkpoints/Prompt_guided_Mask_Decoder.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8771b38a57ad67750dbe18fff95e66c8d1789436096738cace37d58bfc06151e
+size 16310784

modules/pe3r/models.py

@@ -12,17 +12,20 @@ from modules.mobilesamv2 import sam_model_registry
 class Models:
     def __init__(self, device):
         # -- mast3r --
-        MAST3R_CKP = './checkpoints/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric.pth'
+        # MAST3R_CKP = './checkpoints/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric.pth'
+        MAST3R_CKP = 'naver/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric'
         self.mast3r = AsymmetricMASt3R.from_pretrained(MAST3R_CKP).to(device)
 
         # -- sam2 --
-        SAM2_CKP = "./checkpoints/sam2.1_hiera_large.pt"
+        # SAM2_CKP = "./checkpoints/sam2.1_hiera_large.pt"
+        SAM2_CKP = 'facebook/sam2.1-hiera-large'
         SAM2_CONFIG = "./configs/sam2.1/sam2.1_hiera_l.yaml"
         self.sam2 = build_sam2_video_predictor(SAM2_CONFIG, SAM2_CKP, device=device, apply_postprocessing=False)
         self.sam2.eval()
 
         # -- mobilesamv2 & sam1 --
-        SAM1_ENCODER_CKP = './checkpoints/sam_vit_h.pt'
+        # SAM1_ENCODER_CKP = './checkpoints/sam_vit_h.pt'
+        SAM1_ENCODER_CKP = 'facebook/sam-vit-huge'
         SAM1_DECODER_CKP = './checkpoints/Prompt_guided_Mask_Decoder.pt'
         self.mobilesamv2 = sam_model_registry['sam_vit_h'](None)
         image_encoder=sam_model_registry['sam_vit_h_encoder'](SAM1_ENCODER_CKP)