[Feature] Add annotation conversion tool from Labelbee to COCO

tools/dataset/labelbee_to_coco.py

@@ -0,0 +1,154 @@
+import os
+import json
+import shutil
+import argparse
+
+'''
+    将 labelbee多步标注 后的关键点数据进行标签转化，转至mmpose模型需要的coco格式。 数据集先划分好。
+'''
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--input-file', type=str, default='/home/wanghao/Work/labelbee_json_2_MMPose_COCO_json/', help='输入labelbee json的根路径')
+    parser.add_argument('--output-file', type=str, default='/home/wanghao/Work/labelbee_json_2_MMPose_COCO_json/', help='输出mmpose json文件根路径')
+    parser.add_argument('--dataset-split', type=list, default=['train', 'val'], help='数据集划分，对应labelbee根路径下划分情况')
+    args = parser.parse_args()
+
+
+    data_set = args.dataset_split
+    for n in range(len(data_set)):
+        anno_path = args.input_file + data_set[n] + "/"
+        anno_list  = os.listdir(anno_path)
+
+        sort_anno_list = []
+        for i in range(len(anno_list)):
+            sort_anno_list.append(int(anno_list[i].split(".jpg")[0]))
+        sort_anno_list = sorted(sort_anno_list)
+
+        anno_list = []
+        for i in range(len(sort_anno_list)):
+            anno_list.append(str(sort_anno_list[i]) + ".jpg.json")
+        print(anno_list)
+
+        anno_info = {}
+        anno_info["info"] = {"description":"For testing COCO dataset only.",
+                             "year":2022,
+                             "date_created":"2022/03/24"}
+
+        anno_info["licenses"] = [{"url":"http://creativecommons.org/licenses/by-nc-sa/2.0/",
+                                  "id":1,
+                                  "name":"Attribution-NonCommercial-ShareAlike License"}]
+
+        anno_info["categories"] = [{"supercategory":"person",
+                                    "id":1,
+                                    "name":"preson",
+                                    "keypoints":["nose",
+                                                 "left_eye",
+                                                 "right_eye",
+                                                 "left_ear",
+                                                 "right_ear",
+                                                 "left_shoulder",
+                                                 "right_shoulder",
+                                                 "left_elbow",
+                                                 "right_elbow",
+                                                 "left_wrist",
+                                                 "right_wrist",
+                                                 "left_hip",
+                                                 "right_hip",
+                                                 "left_knee",
+                                                 "right_knee",
+                                                 "left_ankle",
+                                                 "right_ankle"],
+                                    "skeleton":[[16, 14],
+                                                [14, 12],
+                                                [17, 15],
+                                                [15, 13],
+                                                [12, 13],
+                                                [6, 12],
+                                                [7, 13],
+                                                [6, 7],
+                                                [6, 8],
+                                                [7, 9],
+                                                [8, 10],
+                                                [9, 11],
+                                                [2, 3],
+                                                [1, 2],
+                                                [1, 3],
+                                                [2, 4],
+                                                [3, 5],
+                                                [4, 6],
+                                                [5, 7]]}]
+
+        anno_info["images"] = []
+
+        anno_info["annotations"] = []
+
+        l = 1
+        for i in range(len(anno_list)):
+            with open(anno_path + anno_list[i], 'r') as f:
+                current_json_content = json.load(f)
+            f.close()
+            img_anno = {
+                        "license":1,
+                        "file_name":anno_list[i].split(".json")[0],
+                        "coco_url":"http://creativecommons.org/licenses/by-nc-sa/2.0/",
+                        "height":current_json_content["height"],
+                        "width":current_json_content["width"],
+                        "date_captured":"2022/03/24",
+                        "flickr_url":"http://creativecommons.org/licenses/by-nc-sa/2.0/",
+                        "id": i + 1
+                        }
+            anno_info["images"].append(img_anno)
+
+            for j in range(len(current_json_content["step_1"]["result"])):
+                current_bbox = current_json_content["step_1"]["result"][j]
+
+                # 凭借这个id在后面的关键点列表里寻找到对应框内的关键点
+                bbox_id = current_bbox["id"]
+                keypoints_list = []
+                keypoints_num = 0
+                area = int(current_bbox["width"]) * int(current_bbox["height"])
+
+                current_json_content["step_2"]["result"]
+                for k in range(len(current_json_content["step_2"]["result"])):
+                    if current_json_content["step_2"]["result"][k]["sourceID"] == bbox_id:
+                        current_bbox_keypoints = current_json_content["step_2"]["result"][k:k+17]
+                        break
+
+                for k in range(len(current_bbox_keypoints)):
+                    if current_bbox_keypoints[k]["attribute"] == "1":
+                        # 这个点 可视
+                        keypoints_num = keypoints_num + 1
+                        keypoints_list.append(int(current_bbox_keypoints[k]["x"]))
+                        keypoints_list.append(int(current_bbox_keypoints[k]["y"]))
+                        keypoints_list.append(2)
+                    else:
+                        # 这个点 不可视
+                        keypoints_list.append(0)
+                        keypoints_list.append(0)
+                        keypoints_list.append(0)
+
+                keypoints_anno = {
+                                    "segmentation":[[]],
+                                    "num_keypoints":keypoints_num,
+                                    "area":area,
+                                    "iscrowd":0,
+                                    "keypoints":keypoints_list,
+                                    "image_id": i + 1,
+                                    "bbox":[int(current_bbox["x"]), int(current_bbox["y"]),
+                                            int(current_bbox["width"]), int(current_bbox["height"])],
+                                    "category_id":1,
+                                    "id": l
+                                  }
+                l = l + 1
+                anno_info["annotations"].append(keypoints_anno)
+
+        with open(args.output_file + data_set[n] +  '.json', 'w') as f:
+            f.write(json.dumps(anno_info, ensure_ascii=False, indent=4, separators=(',', ':')))
+            f.close()
+
+        print(str(data_set[n]) + "标签转化完成！！！！")
+
+
+if __name__=='__main__':
+    main()