Merge pull request #27 from naseemap47/kitti

Kitti

README.md

@@ -431,4 +431,21 @@ to convert mixed images & labels into respective directory of images and labels
 **Example:**
 ```
 python3 tools/create_yolo.py -i path_to/data Dir
+```
+
+### 9. yolo_to_kitti.py:
+to convert YOLO annotations (.txt) to KITTI format
+
+<details>
+  <summary>Args</summary>
+
+  `-i`, `--img` : path to image file <br>
+  `-t`, `--txt` : path to txt file <br>
+  `-c`, `--classes` : path to classes.txt
+
+</details>
+
+**Example:**
+```
+python3 tools/yolo_to_kitti.py -i path_to/image -t path_to/txt -c path_to_classes.txt
 ```

tools/yolo_to_kitti.py

@@ -0,0 +1,72 @@
+import argparse
+import os
+import cv2
+import glob
+
+
+ap = argparse.ArgumentParser()
+ap.add_argument("-i", "--image", type=str, required=True,
+                help="path to image/dir")
+ap.add_argument("-t", "--txt", type=str, required=True,
+                help="path to txt/dir")
+ap.add_argument("-c", "--classes", type=str, required=True,
+                help="path to classes.txt")
+
+args = vars(ap.parse_args())
+path_to_img = args["image"]
+path_to_txt = args['txt']
+path_to_class = args['classes']
+
+os.makedirs("KITTI_"+path_to_txt, exist_ok=True)
+txt_list = sorted(glob.glob(f'{path_to_txt}/*.txt'))
+img_full_list = glob.glob(f'{path_to_img}/*.jpeg') + \
+                glob.glob(f'{path_to_img}/*.jpg')  + \
+                glob.glob(f'{path_to_img}/*.png')
+
+img_list = sorted(img_full_list)
+class_names = open(f'{path_to_class}', 'r+').read().splitlines()
+for txt, img in zip(txt_list, img_list):
+    img_file = cv2.imread(img)
+    image_height, image_width, _ = img_file.shape
+
+    with open(txt, 'r') as fh:
+        data=fh.readlines()
+
+    for cc, lines in enumerate(data):
+        lines = lines.replace('\n', '').split(" ")
+        lines[0] = int(lines[0])
+        lines[1] = float(lines[1])*image_width
+        lines[2] = float(lines[2])*image_height
+        lines[3] = float(lines[3])*image_width
+        lines[4] = float(lines[4])*image_height
+
+        lines[0] = class_names[lines[0]]
+        n_line = [0] * 15
+        n_line[0] = lines[0]
+        n_line[4] = int(lines[1]-lines[3]/2)
+        n_line[5] = int(lines[2]-lines[4]/2)
+        n_line[6] = int(lines[1]+lines[3]/2)
+        n_line[7] = int(lines[2]+lines[4]/2)
+
+        for c, n in enumerate(n_line):
+            if not c == 0:
+                if n < 0:
+                    n_line[c] = 1
+                if n > 256:
+                    n_line[c] = 255
+
+        for cc2, char in enumerate(n_line):
+            if cc2 == 1:
+                n_line[cc2] = float(0)
+            if cc2 > 7:
+                n_line[cc2] = 1
+
+        str1 = ' '.join(str(n_line)).replace(' ', '').replace(',', ' ').replace('[', '').replace("]","").replace("'","").replace('"',"")
+        data[cc] = str1+"\n"
+
+    strf = " ".join(str(x) for x in data)
+    strf = strf.replace("\n ","\n")
+    file = open("KITTI_"+txt, "w")
+    file.write(strf)
+    file.close()
+    print(f"[INFO] Completed {txt}")