BDD100k转coco格式(目标检测)

从这篇文章看到有转换的python代码，但是直接使用会报错：

10%|████████████████                                                                                                                                                   | 6879/69863 [00:00<00:06, 9376.28it/s]
Traceback (most recent call last):
  File "bdd2coco.py", line 112, in <module>
    bdd2coco_detection(attr_id_dict, train_labels, out_fn)
  File "bdd2coco.py", line 45, in bdd2coco_detection
    for label in i['labels']:
KeyError: 'labels'

根据报错信息，可知这是字典的键错误，可能是键值不匹配，输出处理的文件名，发现到 11ecaf4a-837e3550.jpg 就会引起错误，打开json原文件，定位这个文件名的位置，相比于其他的图片，没有了'label'等部分。

{
    "name": "11ecaf4a-837e3550.jpg",
    "attributes": {
      "weather": "clear",
      "timeofday": "night",
      "scene": "highway"
    },
    "timestamp": 10000
},

解决方案

在下载的代码中添加这样两行代码即可（我的思路是直接跳过处理这张图片，相当于舍弃了）：

if 'labels' not in i.keys():
	continue

补充

1. BDD100k类别

根据官方文档，需要将类别和对应id更新为：

1: pedestrian
2: rider
3: car
4: truck
5: bus
6: train
7: motorcycle
8: bicycle
9: traffic light
10: traffic sign

相应的代码修改为：

attr_dict["categories"] = [
	{"supercategory": "none", "id": 1, "name": "pedestrian"},
    {"supercategory": "none", "id": 2, "name": "rider"},
    {"supercategory": "none", "id": 3, "name": "car"},
    {"supercategory": "none", "id": 4, "name": "truck"},
    {"supercategory": "none", "id": 5, "name": "bus"},
    {"supercategory": "none", "id": 6, "name": "train"},
    {"supercategory": "none", "id": 7, "name": "motorcycle"},
    {"supercategory": "none", "id": 8, "name": "bicycle"},
    {"supercategory": "none", "id": 9, "name": "traffic light"},
    {"supercategory": "none", "id": 10, "name": "traffic sign"}
]

2. bdd100k_labels_images_train.json

这个文件就是官网下载的bdd100k_det_20_labels_trainval.zip解压后的train.json，我下载解压以后是det_train.json和det_val.json，代码中的文件目录替换为自己本地这两个文件的目录即可。