ADD CONVERSION SCRIPT

data/scripts/coco-to-yolo.py

@@ -0,0 +1,80 @@
+from pycocotools.coco import COCO
+import cv2
+import numpy as np
+import glob
+import shutil
+
+# Truncates numbers to N decimals
+def truncate(n, decimals=0):
+    multiplier = 10 ** decimals
+    return int(n * multiplier) / multiplier
+
+
+# Download Object 365 from the Object 365 website and put in the same directory as this script
+coco = COCO("instances_train.json")
+cats = coco.loadCats(coco.getCatIds())
+nms = [cat["name"] for cat in cats]
+print("COCO categories: \n{}\n".format(" ".join(nms)))
+cash = set()
+
+for categoryId, cat in enumerate(nms):
+    catIds = coco.getCatIds(catNms=[cat])
+    imgIds = coco.getImgIds(catIds=catIds)
+    images = coco.loadImgs(imgIds)
+    print(cat)
+    # # Create a subfolder in this directory called "labels". This is where the annotations will be saved in YOLO format
+    for im in images:
+        dw = 1.0 / im["width"]
+        dh = 1.0 / im["height"]
+
+        annIds = coco.getAnnIds(imgIds=im["id"], catIds=catIds, iscrowd=None)
+        anns = coco.loadAnns(annIds)
+
+        path = im["file_name"].split("/")
+        fixed_path = path[-1]
+        filename = fixed_path.replace(".jpg", ".txt")
+
+        try:
+            # Test image for corruption
+            if fixed_path not in cash:
+                img = cv2.imread(f"images/train/{fixed_path}")
+                img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+                cash.add(fixed_path)
+
+            with open("/labels/train/" + filename, "a+") as myfile:
+                for i in range(len(anns)):
+                    xmin = anns[i]["bbox"][0]
+                    ymin = anns[i]["bbox"][1]
+                    xmax = anns[i]["bbox"][2] + anns[i]["bbox"][0]
+                    ymax = anns[i]["bbox"][3] + anns[i]["bbox"][1]
+
+                    x = (xmin + xmax) / 2
+                    y = (ymin + ymax) / 2
+
+                    w = xmax - xmin
+                    h = ymax - ymin
+
+                    x = x * dw
+                    w = w * dw
+                    y = y * dh
+                    h = h * dh
+
+                    # Note: This assumes a single-category dataset, and thus the "0" at the beginning of each line.
+                    mystring = f"{categoryId} {truncate(x, 7)} {truncate(y, 7)} {truncate(w, 7)} {truncate(h, 7)}"
+                    # mystring = str(categoryId + " " + str(truncate(x, 7)) + " " + str(truncate(y, 7)) + " " + str(truncate(w, 7)) + " " + str(truncate(h, 7)))
+                    myfile.write(mystring)
+                    myfile.write("\n")
+            myfile.close()
+
+        except Exception as e:
+            print(e)
+
+current_dir = "images/train"
+myset = {2, 25, 4, 5}
+for fullpath in glob.iglob(os.path.join(current_dir, "*.jpg")):
+    n = random.randint(1, 100)
+    if n in myset:
+        title, ext = os.path.splitext(os.path.basename(fullpath))
+        print(title)
+        shutil.move(f"../images/train/{title}.jpg", f"../images/val/{title}.jpg")
+        shutil.move(f"../labels/train/{title}.txt", f"../labels/val/{title}.txt")