Split car bus truck

class_list.txt

@@ -1,4 +1,12 @@
-closed_door
-opened_door
-bus
-number
+Person
+Bike
+Car
+Small Animal
+Large Animal
+Bird
+Bus
+Truck
+Micro‑vessel
+Small vessel
+Medium-size vessel
+Large vessel  

refactor.py

@@ -0,0 +1,17 @@
+import os
+
+bb_dir = "new_bbox_txt/"
+
+
+if __name__ == "__main__":
+    c = 0
+    for f in os.listdir(bb_dir):
+        if os.path.getsize(bb_dir + f) != 0:
+            with open(bb_dir + f, "r+") as f:
+                elems = f.readline().split()
+                while elems:
+                    if elems and len(elems) != 5:
+                        f.truncate(0)
+                        c += 1
+                    elems = f.readline().split()
+    print(c)

relabel_car_ai.py

@@ -0,0 +1,133 @@
+import os
+from PIL import Image
+from transformers import pipeline
+
+DATASET_ROOT = "D:/VS_code/Praktika/ModifiedOpenLabelling"
+
+IMAGES_ROOT = os.path.join(DATASET_ROOT, "images/train")
+LABELS_ROOT = os.path.join(DATASET_ROOT, "labels/train")
+OUTPUT_LABELS_ROOT = os.path.join(DATASET_ROOT, "labels_new/train")
+
+OLD_CAR_LABEL = 2
+
+NEW_LABELS = {
+    "car": 2,
+    "bus": 6,
+    "truck": 7
+}
+
+MILITARY_KEYWORDS = [
+    "tank",
+    "armored",
+    "armoured",
+    "military",
+    "apc",
+    "artillery",
+    "missile",
+    "launcher",
+    "howitzer"
+]
+
+IMAGE_EXTENSIONS = (".jpg", ".jpeg", ".png")
+
+classifier = pipeline(
+    "image-classification",
+    model="google/vit-base-patch16-224"
+)
+
+def yolo_to_pixels(bbox, img_w, img_h):
+    x_c, y_c, w, h = bbox
+    x1 = int((x_c - w / 2) * img_w)
+    y1 = int((y_c - h / 2) * img_h)
+    x2 = int((x_c + w / 2) * img_w)
+    y2 = int((y_c + h / 2) * img_h)
+    return max(0, x1), max(0, y1), min(img_w, x2), min(img_h, y2)
+
+def classify_crop(crop):
+    preds = classifier(crop)
+
+    for p in preds:
+        label = p["label"].lower()
+
+        for kw in MILITARY_KEYWORDS:
+            if kw in label:
+                return "truck"
+
+        if "bus" in label:
+            return "bus"
+
+        if "truck" in label or "lorry" in label:
+            return "truck"
+
+        if "car" in label:
+            return "car"
+
+    return "truck"
+
+for root, _, files in os.walk(IMAGES_ROOT):
+    for file in files:
+        if not file.lower().endswith(IMAGE_EXTENSIONS):
+            continue
+
+        image_path = os.path.join(root, file)
+
+        rel_path = os.path.relpath(image_path, IMAGES_ROOT)
+        rel_dir = os.path.dirname(rel_path)
+
+        label_path = os.path.join(
+            LABELS_ROOT,
+            rel_dir,
+            os.path.splitext(file)[0] + ".txt"
+        )
+
+        if not os.path.exists(label_path):
+            continue
+
+        image = Image.open(image_path).convert("RGB")
+        img_w, img_h = image.size
+
+        with open(label_path, "r") as f:
+            lines = f.readlines()
+
+        new_lines = []
+
+        for line in lines:
+            line_strip = line.strip()
+            if not line_strip:
+                continue
+            parts = line.strip().split()
+            cls = int(parts[0])
+            bbox = list(map(float, parts[1:]))
+
+            if cls != OLD_CAR_LABEL:
+                new_lines.append(line.strip())
+                continue
+
+            x1, y1, x2, y2 = yolo_to_pixels(bbox, img_w, img_h)
+            crop = image.crop((x1, y1, x2, y2))
+
+            new_class_name = classify_crop(crop)
+            new_class_id = NEW_LABELS[new_class_name]
+
+            new_line = " ".join(
+                [str(new_class_id)] + [f"{v:.6f}" for v in bbox]
+            )
+            new_lines.append(new_line)
+
+            print(new_class_name, "←", rel_path)
+
+        output_dir = os.path.join(OUTPUT_LABELS_ROOT, rel_dir)
+        os.makedirs(output_dir, exist_ok=True)
+
+        output_label_path = os.path.join(
+            output_dir,
+            os.path.splitext(file)[0] + ".txt"
+        )
+
+        with open(output_label_path, "w") as f:
+            if new_lines:
+                f.write("\n".join(new_lines) + "\n")
+            else:
+                f.write("\n")
+
+        print(f"✔ processed {rel_path}")

rename.py

@@ -0,0 +1,40 @@
+import os
+import shutil
+
+
+bb_dir = "bbox_txt/"
+new_bb_dir = "new_bbox_txt/"
+
+if os.path.isdir(new_bb_dir) == False:
+    os.makedirs(new_bb_dir)
+
+
+def is_wrong(txt_path):
+    return txt_path.count('_txt') != 0
+
+
+def png_in_name(txt_path):
+    return txt_path.count('_png.') != 0 or txt_path.count('_PNG.')
+
+
+if __name__ == "__main__":
+    for f in os.listdir(bb_dir):
+        f_path = os.path.join(bb_dir, f)
+        new_f_path = f.replace('_txt', '_jpg')
+        print(f_path, new_f_path)
+        if os.path.getsize(f_path) != 0 and is_wrong(f):
+            new_f_path = os.path.join(new_bb_dir, new_f_path)
+            shutil.copy(f_path, new_f_path)
+            print("The file name has been changed: old file path - " + f_path
+                  + " | new file path " + new_f_path)
+        elif os.path.getsize(f_path) == 0 and is_wrong(f):
+            right_f_path = os.path.join(bb_dir, new_f_path)
+            new_f_path = os.path.join(new_bb_dir, new_f_path)
+            shutil.copy(right_f_path, new_f_path)
+            print("The file name has been changed: old file path - " + right_f_path
+                  + " | new file path " + new_f_path)
+        elif png_in_name(f):
+            new_f_path = os.path.join(new_bb_dir, f)
+            shutil.copy(f_path, new_f_path)
+            print("The file name has been changed: old file path - " + f_path
+                  + " | new file path " + new_f_path)

requirements.txt

@@ -1,2 +1,3 @@
 opencv-python
-numpy
+numpy
+natsort