v1.1

Author: SelimSavas

Diff for: SAMandYOLOv8video.py

@@ -0,0 +1,133 @@
+!pip install ultralytics
+!pip install supervision
+!pip install roboflow
+!{sys.executable} -m pip install 'git+https://github.com/facebookresearch/segment-anything.git'
+!wget -q https://dl.fbaipublicfiles.com/segment_anything/sam_vit_l_0b3195.pth
+
+import numpy as np
+import cv2
+import os
+import sys
+import torch
+
+#ROBOFLOW
+from roboflow import Roboflow
+rf = Roboflow(api_key="Pho4XVVaJnNcR3y3HNrK")
+project = rf.workspace("team12").project("airplane-bp8fl")
+dataset = project.version(2).download("yolov8")
+
+# YOLO
+from ultralytics import YOLO
+yolo_model = YOLO('yolov8x.pt')
+
+# SEGMENT-Anything
+from segment_anything import sam_model_registry, SamAutomaticMaskGenerator, SamPredictor
+import supervision as sv
+
+DEVICE = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+MODEL_TYPE = "vit_l"
+
+CHECKPOINT_PATH = os.path.join("sam_vit_l_0b3195.pth")
+sam = sam_model_registry[MODEL_TYPE](checkpoint=CHECKPOINT_PATH).to(device=DEVICE)
+
+mask_predictor = SamPredictor(sam)
+
+# Run Code
+
+VIDEO_SOURCE = '/content/airplane2.mp4'
+VIDEO_DESTINATION = '/content/result2.avi'
+
+# Video kaynağını aç
+cap = cv2.VideoCapture(VIDEO_SOURCE)
+
+# Video kaydedicisini ayarla
+width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+fourcc = cv2.VideoWriter_fourcc(*'XVID')
+FPS = 30.0
+out = cv2.VideoWriter(VIDEO_DESTINATION, fourcc, FPS, (width, height))
+
+while True:
+    # Frame oku
+    ret, frame = cap.read()
+
+    # Okuma başarısızsa video bitti
+    if not ret:
+        break
+
+    # Girdi frame'ini SuperAnnotate'e uygun formata getir
+    frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+    results = yolo_model.predict(source=frame)
+    
+    boxes_class_name = np.array([])
+    for r in results:
+        for c in r.boxes.cls:
+            boxes_class_name = np.append(boxes_class_name,yolo_model.names[int(c)])
+    boxes_class_name
+
+    k = 0
+    result_np_son = np.array([])
+    result_class = np.array([])
+    result_score = np.array([])
+    for result in results:
+        result_np = np.array(result.boxes.data.cpu())
+
+    for i in result_np:
+        result_np_son = np.append(result_np_son,i[:][:-2])
+        result_class = np.append(result_class,i[:][-1:])
+        result_score = np.append(result_score,i[:][-2:-1])
+        k += 1
+
+
+    boxes = result_np_son.reshape((k,4))
+    boxes_class = result_class.reshape((k,))
+    boxes_score = result_score.reshape((k,))
+
+
+    segmented_image = frame.copy()
+    box_image = frame.copy()
+    boxes_class_value = {'0':sv.Color.white(),
+                         '1':sv.Color.blue(), 
+                         '2':sv.Color.blue(),
+                         '3':sv.Color.blue(),
+                         '4':sv.Color.red(),
+                         '5':sv.Color.blue(),
+                         '7':sv.Color.black(),
+                         '8':sv.Color.green(),
+                         '9':sv.Color.blue(),
+                         '37':sv.Color.blue()}
+    mask_predictor.set_image(frame)
+
+    boxes = result.boxes.xyxy
+    if len(boxes) != 0:
+      masks, scores, logits = mask_predictor.predict_torch(
+          point_coords = None,
+          point_labels = None,
+          boxes = boxes * 1040/max(frame.shape),
+          multimask_output=True
+      )
+
+      masks = masks.cpu().numpy()
+
+      i = 0
+      for box in boxes.cpu().numpy():
+
+          box_annotator = sv.BoxAnnotator(color= boxes_class_value[str(int(boxes_class[i]))], text_padding=10)
+          mask_annotator = sv.MaskAnnotator(color= boxes_class_value[str(int(boxes_class[i]))])
+
+          detections = sv.Detections(
+              xyxy=sv.mask_to_xyxy(masks=masks[i]),
+              mask=masks[i]
+          )
+          
+          detections = detections[detections.area == np.max(detections.area)]
+          box_image = box_annotator.annotate(scene=box_image, detections=detections, skip_label=True)
+          segmented_image = mask_annotator.annotate(scene=segmented_image, detections=detections)
+          i += 1   
+
+    # Video kaydediciye yaz
+    out.write(cv2.cvtColor(segmented_image, cv2.COLOR_RGB2BGR))
+
+# Video kaynak ve kaydediciyi serbest bırak
+cap.release()
+