Rename rest (4).py to app.py

app.py

@@ -0,0 +1,40 @@
+#best restoration model
+
+import gradio as gr
+import numpy as np
+from ultralytics import YOLO
+import cv2
+
+model = YOLO("best2.pt")
+
+with gr.Blocks() as demo:
+    with gr.Row():
+      gr.Markdown("## Web app that segement the roof from house image using YLOV8")
+    with gr.Row():
+      gr.Markdown("### Please upload an image of the house, then click on Segement .")
+    with gr.Row():  
+      gr.Markdown("Then the yolov8 will segement the roof.")
+
+
+    with gr.Row():
+        input_image = gr.Image(source='upload', elem_id="input_image_upload", label="UploadImage")
+        output_image = gr.Image(label="output")
+    with gr.Row():
+        segement_an = gr.Button("Segement")
+    def segement_(input_):
+        
+        img = model.predict(source=input_,
+                         stream=True, retina_masks=True)
+        for result in img:
+            mask = result.masks.cpu().numpy()
+            masks = mask.masks.astype(bool)
+            new = np.ones_like(input_, dtype=np.uint8)
+            for m in masks:
+                new[m] = input_[m]
+            return new
+
+    segement_an.click(segement_, inputs=[input_image], outputs=[output_image])
+    demo.launch(share=True)
+
+
+       

rest (4).py

@@ -1,77 +0,0 @@
-#best restoration model
-
-import gradio as gr
-import numpy as np
-import torch
-from ultralytics import YOLO
-import cv2
-from segment_anything import sam_model_registry, SamAutomaticMaskGenerator, SamPredictor
-import torch
-import torchvision
-
-device = "cuda"
-sam = sam_model_registry["vit_h"](checkpoint="sam_vit_h_4b8939.pth").to(device=torch.device('cuda:0'))
-mask_predictor = SamPredictor(sam)
-model = YOLO("/content/window_model.pt")
-
-with gr.Blocks() as demo:
-    with gr.Row():
-      gr.Markdown("## Web app that segement the windows from house image using YLOV8 and SAM")
-    with gr.Row():
-      gr.Markdown("### Please upload an image of the house, then click on Detect the bounding box.")
-    with gr.Row():  
-      gr.Markdown("Then the yolov8 will detect the bounding boxes of the windows, and feed them to SAM.")
-    with gr.Row():  
-      gr.Markdown("Note:it is not important if the YOLOV8 detect the wrong item,what is important if SAM segement the detect object very well.")
-    with gr.Row():  
-      gr.Markdown("I will imporve the accuracy of YOLOV8 if SAM perform well.")
-
-    with gr.Row():
-        input_image = gr.Image(source='upload', elem_id="input_image_upload", label="UploadImage")
-        detected_output = gr.Image(label="detected windowss")
-        output_image = gr.Image(label="output")
-    with gr.Row():
-        detect_bx = gr.Button("Detect the bounding boxes")
-        segement_an = gr.Button("Segement")
-    def detect_(input_):
-        
-        results = model.predict(input_,conf=0.25)
-        predicted_boxes = results[0].boxes.xyxy
-        for result in predicted_boxes:
-            new = np.ones_like(input_, dtype=np.uint8)
-            mask = result.masks.cpu().numpy()
-            masks = mask.masks.astype(bool)
-            for m in masks:
-                new[m] = input_[m]
-        return new 
-    def segement_anythings(input_):
-      predicted_boxes = predict_(input_)
-      transformed_boxes = mask_predictor.transform.apply_boxes_torch(predicted_boxes, input_.shape[:2])
-      mask_predictor.set_image(input_)
-      masks, scores, logits = mask_predictor.predict_torch(
-        boxes = transformed_boxes,
-        multimask_output=False,
-        point_coords=None,
-        point_labels=None
-      )
-      new = np.ones_like(input_, dtype=np.uint8)
-      for mask in masks:
-        k = mask.cpu().numpy()[0]
-        new[k] = input_[k]
-      return new
-    def predict_(input_):
-      results = model.predict(source=input_, conf=0.25)
-      predicted_boxes = results[0].boxes.xyxy
-      return predicted_boxes
-    def detect_b(input_):
-      predicted_boxes = predict_(input_)
-      for box in predicted_boxes:
-         cv2.rectangle(input_, (int(box[0]), int(box[1])), (int(box[2]), int(box[3])), (0, 255, 0), 2)
-      return input_
-
-    detect_bx.click(detect_b, inputs=[input_image], outputs=[detected_output])
-    segement_an.click(segement_anythings, inputs=[input_image], outputs=[output_image])
-    demo.launch(share=True)
-
-
-