debug

examples/blobctrl/blobctrl_app.py

@@ -114,7 +114,7 @@ blobnet = BlobNetModel.from_pretrained(blobnet_path, ignore_mismatched_sizes=Tru
 ## sam
 print(f"Loading SAM...")
 mobile_sam = sam_model_registry['vit_h'](checkpoint=sam_path).to(device)
-mobile_sam.eval()
+# mobile_sam.eval()
 mobile_predictor = SamPredictor(mobile_sam)
 colors = [(255, 0, 0), (0, 255, 0)]
 markers = [1, 5]
@@ -1030,14 +1030,17 @@ def segmentation(img, sel_pix):
     mobile_predictor.set_image(img if isinstance(img, np.ndarray) else np.array(img))
     with torch.no_grad():
         masks, _, _ = mobile_predictor.predict(point_coords=np.array(points), point_labels=np.array(labels), multimask_output=False)
-    print("================")
+    print("=======img=========")
     print(img)
     print(img.shape)
+    print("=======points and labels=========")
     print(points)
     print(labels)
+    print("=======masks=========")
     print(masks)
     print(np.unique(masks))
     print("================")
+    print(mobile_predictor)
     output_mask = np.ones((masks.shape[1], masks.shape[2], 3))*255
     for i in range(3):
         output_mask[masks[0] == True, i] = 0.0
@@ -1059,11 +1062,11 @@ def get_point(img, sel_pix, evt: gr.SelectData):
     # online show seg mask
     print(evt.index)
     masked_img, output_mask = segmentation(img, sel_pix)
-    print(masked_img.shape)
-    print(output_mask.shape)
-    print(masked_img)
-    print(output_mask)
-    print(np.unique(output_mask))
+    # print(masked_img.shape)
+    # print(output_mask.shape)
+    # print(masked_img)
+    # print(output_mask)
+    # print(np.unique(output_mask))
 
     return masked_img.astype(np.uint8), output_mask
 

requirements.txt

@@ -4,10 +4,10 @@ accelerate==1.5.2
 huggingface_hub==0.29.3
 gradio==5.21.0
 opencv-python==4.8.1.78
-numpy==1.26.0
+numpy==1.26.2
 einops==0.8.1
 matplotlib==3.10.1
-segment_anything
+segment_anything==1.0
 torch==2.2.0
 torchvision==0.17.0
 torchaudio==2.2.0