first commit

get_fine_dataset.py

@@ -53,7 +53,7 @@ def VocToFineWithLabel(num_count,root_path,image_T):
             ymin = object_id.getElementsByTagName('bndbox')[0].getElementsByTagName('ymin')[0].childNodes[0].data
             xmax = object_id.getElementsByTagName('bndbox')[0].getElementsByTagName('xmax')[0].childNodes[0].data
             ymax = object_id.getElementsByTagName('bndbox')[0].getElementsByTagName('ymax')[0].childNodes[0].data
-            name = object_id.getElementsByTagName('name')[0].childNodes[0].data
+            name = "bird"#object_id.getElementsByTagName('name')[0].childNodes[0].data
             ratio_x = (3840.0/3840.0)
             ratio_y = (2160.0/2160.0)
             #print(ratio_x,"    ",ratio_y)
@@ -65,7 +65,10 @@ def VocToFineWithLabel(num_count,root_path,image_T):
             ymax = int(float(ymax)*ratio_y)
             if ((xmax - xmin)<image_T) | ((ymax - ymin)<image_T):
                 continue
-            img1 = cv2_crop(img,(xmin, ymin,xmax, ymax))
+            try:
+                img1 = cv2_crop(img,(xmin, ymin,xmax, ymax))
+            except:
+                print(xml_file)
             if not os.path.exists(os.path.join(root_path.replace("Annotations","FineDataset_{}".format(str(image_T))),str(name).replace("\t","").strip())):
                 os.makedirs(os.path.join(root_path.replace("Annotations","FineDataset_{}".format(str(image_T))),str(name).replace("\t","").strip()))
             cv2.imwrite(os.path.join(root_path.replace("Annotations","FineDataset_{}".format(str(image_T))),str(name).replace("\t","").strip(),"{}.jpg".format(str(num_count).zfill(5))), img1)
@@ -84,9 +87,9 @@ if __name__ == '__main__':
     parser1.add_argument("image_T", default='0001', type=str,
                         help='Video name.')
     parser1.add_argument("root_path", default='5', type=str,
-                        help='GPU.')
+                        help='Annotations')
     args = parser1.parse_args()
     num_count = 0
-    VocToFineWithLabel(num_count,args.root_path, int(args.image_T))
+    VocToFineWithLabel(num_count, args.root_path, int(args.image_T))