YOLOv5 → COCO | Label Convert Documentation

[giscus[bot]]

YOLOv5 → COCO | Label Convert Documentation

简介 link将YOLOv5格式数据集转换为COCO格式。
支持标注格式为矩形框和多边形框。
YOLOv5数据结构如下 link notifications 具体结构示例文件，可移步：yolov5_dataset
yolov5_dataset ├── classes.txt ├── non_labels # 通常用来放负样本 │ └── bg1.jpeg ├── images │ ├── images(13).jpg │ └── images(3).jpg ├── labels │ ├── images(13).txt │ └── images(3).txt ├── train.txt └── val.txt 转换 link yolov5_to_coco --data_dir dataset/yolov5_dataset --mode_list train,val --data_dir: 数据集所在目录。示例为dataset/yolov5_dataset --save_dir: 保存转换后的数据集目录。默认为dataset/yolov5_dataset_coco --mode_list: 指定划分的数据集种类。 (例如：train,val,test / train,val) 转换后结构如下： link notifications 具体结构示例文件，可移步：COCO_dataset
COCO_dataset ├── annotations │ ├── instances_train2017.json │ └── instances_val2017.json ├── train2017 │ ├── 000000000001.

https://rapidai.github.io/LabelConvert/docs/supportconversions/yolov5_to_coco/

[limingxuan12]

The dataset format marked on MaixHub is in XML format, with images in one folder and training and validation files in another folder. Is this type of format convertible?

[SWHL]

Could you provide details?

[production23]

Hi @limingxuan12,

We’ve seen a lot of teams run into the same situation, datasets living in different folder/format structures (XML, YOLO, COCO, etc.) and needing them standardized for training. It’s definitely doable.

If you’re interested, we’ve set up pilots where we handle this kind of conversion + validation in one workflow so you don’t have to wrangle formats by hand. Happy to chat if that would help your project.

Best, Aram @ EvoLearns