COCO2017

## 数据集描述
COCO2017数据集，[官方地址](http://cocodataset.org/)。

### 数据集简介
COCO2017包含训练集118287张，验证集5000张，80类。此外，该仓库还提供了一个训练子集（25504张）。

### 数据集支持的任务
实例分割。

## 数据集的格式和结构

### 数据格式

目录结构
```
COCO2017train
├── annotations
│ └── instances_train2017.json
└── train2017
COCO2017subtrain
├── annotations
│ └── instances_subtrain2017.json
└── subtrain2017
COCO2017val
├── annotations
│ └── instances_val2017.json
└── val2017
```

COCO的实例分割标注字段如下，完整的细节可参考[这里](https://cocodataset.org/#format-data)：
```
{
"images": [image],
"annotations": [annotation],
"categories": [category]
}
image = {
"id": int,
"width": int,
"height": int,
"file_name": str,
}
annotation = {
"id": int,
"image_id": int,
"category_id": int,
"segmentation": RLE or [polygon],
"area": float,
"bbox": [x,y,width,height],
"iscrowd": 0 or 1,
}
categories = [{
"id": int,
"name": str,
"supercategory": str,
}]
```

### 数据集加载方式
本仓库目前提供了训练集、验证集和一个训练子集的下载。
```python
from modelscope.msdatasets import MsDataset
from modelscope.utils.constant import DownloadMode
dataset_train = MsDataset.load('COCO2017_Instance_Segmentation', split='train',
download_mode=DownloadMode.FORCE_REDOWNLOAD)
dataset_subtrain = MsDataset.load('COCO2017_Instance_Segmentation', split='subtrain',
download_mode=DownloadMode.FORCE_REDOWNLOAD)
dataset_val = MsDataset.load('COCO2017_Instance_Segmentation', split='validation',
download_mode=DownloadMode.FORCE_REDOWNLOAD)
print(dataset_train.config_kwargs)
print(dataset_subtrain.config_kwargs)
print(dataset_val.config_kwargs)
```

## 数据集版权信息
本数据集遵循Creative Commons Attribution 4.0 License，更多的版权、授权使用信息请参考[这里](https://cocodataset.org/#termsofuse)。

## 引用方式

```
@inproceedings{lin2014microsoft,
title={Microsoft coco: Common objects in context},
author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence},
booktitle={European conference on computer vision},
pages={740--755},
year={2014},
organization={Springer}
}
```

## Dataset Description COCO2017 dataset, [official website](http://cocodataset.org/). ### Dataset Introduction COCO2017 contains 118,287 training images, 5,000 validation images, and 80 categories. In addition, this repository provides a training subset (25,504 images). ### Tasks Supported by the Dataset Instance segmentation. ## Dataset Format and Structure ### Data Format Directory structure ``` COCO2017train ├── annotations │ └── instances_train2017.json └── train2017 COCO2017subtrain ├── annotations │ └── instances_subtrain2017.json └── subtrain2017 COCO2017val ├── annotations │ └── instances_val2017.json └── val2017 ``` The instance segmentation annotation fields of COCO are as follows. For complete details, please refer to [here](https://cocodataset.org/#format-data): ``` { "images": [image], "annotations": [annotation], "categories": [category] } image = { "id": int, "width": int, "height": int, "file_name": str, } annotation = { "id": int, "image_id": int, "category_id": int, "segmentation": RLE or [polygon], "area": float, "bbox": [x,y,width,height], "iscrowd": 0 or 1, } categories = [{ "id": int, "name": str, "supercategory": str, }] ``` ### Dataset Loading Method This repository currently provides download of the training set, validation set, and a training subset. ```python from modelscope.msdatasets import MsDataset from modelscope.utils.constant import DownloadMode dataset_train = MsDataset.load('COCO2017_Instance_Segmentation', split='train', download_mode=DownloadMode.FORCE_REDOWNLOAD) dataset_subtrain = MsDataset.load('COCO2017_Instance_Segmentation', split='subtrain', download_mode=DownloadMode.FORCE_REDOWNLOAD) dataset_val = MsDataset.load('COCO2017_Instance_Segmentation', split='validation', download_mode=DownloadMode.FORCE_REDOWNLOAD) print(dataset_train.config_kwargs) print(dataset_subtrain.config_kwargs) print(dataset_val.config_kwargs) ``` ## Dataset License Information This dataset is licensed under the Creative Commons Attribution 4.0 License. For more information on copyright and usage permissions, please refer to [here](https://cocodataset.org/#termsofuse). ## Citation ``` @inproceedings{lin2014microsoft, title={Microsoft coco: Common objects in context}, author={Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C Lawrence}, booktitle={European conference on computer vision}, pages={740--755}, year={2014}, organization={Springer} } ```

搜集汇总

背景与挑战

背景概述

COCO2017是一个大型计算机视觉数据集，主要用于实例分割任务，包含超过12万张彩色图像，覆盖80个常见物体类别。数据集提供训练集、验证集和一个训练子集，采用结构化目录和JSON注释格式，便于模型训练和评估。该数据集遵循Apache 2.0和CC BY 4.0许可证，是目标检测和分割领域的重要基准资源。