Upload app.py

app.py

@@ -1,3 +1,4 @@
+import os
 import spaces
 import gradio as gr
 import torch
@@ -6,12 +7,20 @@ from PIL import Image
 from transformers import Sam3Processor, Sam3Model
 import requests
 import warnings
+from huggingface_hub import login, list_repo_files
+
 warnings.filterwarnings("ignore")
 
+HF_TOKEN = os.environ.get("HF_TOKEN")
+login(token=HF_TOKEN)
+
+
 # Global model and processor
 device = "cuda" if torch.cuda.is_available() else "cpu"
-model = Sam3Model.from_pretrained("facebook/sam3", torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32).to(device)
-processor = Sam3Processor.from_pretrained("facebook/sam3")
+model = Sam3Model.from_pretrained("giginho83/sa3-base", token=HF_TOKEN, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32).to(device)
+processor = Sam3Processor.from_pretrained("giginho83/sa3-base", token=HF_TOKEN)
+
+print("Model loaded successfully!")
 
 @spaces.GPU()
 def segment(image: Image.Image, text: str, threshold: float, mask_threshold: float):
@@ -56,7 +65,8 @@ def segment(image: Image.Image, text: str, threshold: float, mask_threshold: flo
             annotations.append((mask_np, label))
         
         scores_text = ", ".join([f"{s:.2f}" for s in results['scores'].cpu().numpy()[:5]])
-        info = f"✅ Found **{n_masks}** objects matching **'{text}'**\nConfidence scores: {scores_text}{'...' if n_masks > 5 else ''}"
+        #info = f"✅ Found **{n_masks}** objects matching **'{text}'**\nConfidence scores: {scores_text}{'...' if n_masks > 5 else ''}"
+        info = f"✅ Detected objects: **{n_masks}"
         
         # Return tuple: (base_image, list_of_annotations)
         return (image, annotations), info
@@ -78,20 +88,10 @@ def segment_example(image_path: str, prompt: str):
 
 # Gradio Interface
 with gr.Blocks(
-    theme=gr.themes.Soft(),
-    title="SAM3 - Promptable Concept Segmentation",
-    css=".gradio-container {max-width: 1400px !important;}"
+    #theme=gr.themes.Soft(),
+    title="Count items in supermarket",
+    #css=".gradio-container {max-width: 1400px !important;}"
 ) as demo:
-    gr.Markdown(
-        """
-        # SAM3 - Promptable Concept Segmentation (PCS)
-        
-        **SAM3** performs zero-shot instance segmentation using natural language prompts.
-        Upload an image, enter a text prompt (e.g., "person", "car", "dog"), and get segmentation masks.
-        
-        Built with [anycoder](https://huggingface.co/spaces/akhaliq/anycoder)
-        """
-    )
     
     gr.Markdown("### Inputs")
     with gr.Row(variant="panel"):
@@ -104,13 +104,13 @@ with gr.Blocks(
         image_output = gr.AnnotatedImage(
             label="Output (Segmented Image)",
             height=400,
-            show_legend=True,
+            show_legend=False,
         )
     
     with gr.Row():
         text_input = gr.Textbox(
             label="Text Prompt",
-            placeholder="e.g., person, ear, cat, bicycle...",
+            placeholder="e.g., bottle box, jar, tube...",
             scale=3
         )
         clear_btn = gr.Button("🔍 Clear", size="sm", variant="secondary")
@@ -140,15 +140,15 @@ with gr.Blocks(
     
     segment_btn = gr.Button("🎯 Segment", variant="primary", size="lg")
     
-    gr.Examples(
-        examples=[
-            ["http://images.cocodataset.org/val2017/000000077595.jpg", "cat"],
-        ],
-        inputs=[image_input, text_input],
-        outputs=[image_output, info_output],
-        fn=segment_example,
-        cache_examples=False,
-    )
+    # gr.Examples(
+    #     examples=[
+    #         ["http://images.cocodataset.org/val2017/000000077595.jpg", "cat"],
+    #     ],
+    #     inputs=[image_input, text_input],
+    #     outputs=[image_output, info_output],
+    #     fn=segment_example,
+    #     cache_examples=False,
+    # )
     
     clear_btn.click(
         fn=clear_all,
@@ -164,7 +164,7 @@ with gr.Blocks(
     gr.Markdown(
         """
         ### Notes
-        - **Model**: [facebook/sam3](https://huggingface.co/facebook/sam3)
+        - **Model**: [giginho83/sa3_base]
         - Click on segments in the output to see labels
         - GPU recommended for faster inference
         """