COCO数据集全称为Microsoft Common Objects in Context(MS COCO)，它是一个大规模(large-scale)的对象检测(object detection)、分割(segmentation)、关键点检测(key-point detection)和字幕(captioning)数据集。此数据集由32.8万张图像组成，官网为：https://cocodataset.org/#home ，论文《Microsoft COCO: Common Objects in Context》：https://arxiv.org/pdf/1405.0312.pdf
COCO数据集的第一个版本于2014年发布，它包含16.4万张图像，分为训练集(8.3万张)、验证集(4.1万张)和测试集(4.1万张)。2015年发布了额外的8.1万张图像测试集，包括所有以前的测试图像和4万张新图像。2017年将训练集/验证集分配从8.3万/4.1万更改为11.8万/0.5万张，新的拆分使用相同的图像和标注(annotation)。2017年测试集是2015年测试集的子集包含4.1万张。此外，2017版本包含一个新的未标注的12.3万张数据集。近几年的Tasks使用的都是2017年的数据集。可以从https://cocodataset.org/#download 直接下载需要的COCO数据集。
COCO API可以帮助加载、解析和可视化COCO中的标注。API支持多种标注格式(annotation formats)。
COCO数据集features：
(1).对象检测：具有80个对象类别(object categories)的边界框(bounding boxes)和每个实例的分割掩码。80个类别包括：person, bicycle, car, motorcycle, airplane, bus, train, truck, boat, traffic light, fire hydrant, stop sign, parking meter, bench, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe, backpack, umbrella, handbag, tie, suitcase, frisbee, skis, snowboard, sports_ball, kite, baseball bat, baseball glove, skateboard, surfboard, tennis racket, bottle, wine glass, cup, fork, knife, spoon, bowl, banana, apple, sandwich, orange, broccoli, carrot, hot_dog, pizza, donut, cake, chair, couch, potted plant, bed, dining table, toilet, tv, laptop, mouse, remote, keyboard, cell phone, microwave, oven, toaster, sink, refrigerator, book, clock, vase, scissors, teddy bear, hair drier, toothbrush。
(2).字幕(captioning)：图像的自然语言描述。
(3).关键点检测：包含超过20万张图像和25万个关键点标记的人物实例(17个可能的关键点：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)。
(4).图像分割(stuff image segmentation)：分为91种类别(stuff categories)，包括：banner, blanket, branch, bridge, building-other, bush, cabinet, cage, cardboard, carpet, ceiling-other, ceiling-tile, cloth, clothes, clouds, counter, cupboard, curtain, desk-stuff, dirt, door-stuff, fence, floor-marble, floor-other, floor-stone, floor-tile, floor-wood, flower, fog, food-other, fruit, furniture-other, grass, gravel, ground-other, hill, house, leaves, light, mat, metal, mirror-stuff, moss, mountain, mud, napkin, net, paper, pavement, pillow, plant-other, plastic, platform, playingfield, railing, railroad, river, road, rock, roof, rug, salad, sand, sea, shelf, sky-other, skyscraper, snow, solid-other, stairs, stone, straw, structural-other, table, tent, textile-other, towel, tree, vegetable, wall-brick, wall-concrete, wall-other, wall-panel, wall-stone, wall-tile, wall-wood, water-other, waterdrops, window-blind, window-other, wood。
(5).全场景分割(panoptic:full scene segmentation)：分为80种类别(thing classes，例如人、自行车、大象)和91种stuff类别(stuff classes，例如草、天空、道路)。1(不属于任何其它类，Id为0，label name为unlabeled)+80+91类别具体Id和对应的Label name参考：https://github.com/nightrome/cocostuff/blob/master/labels.md 。
(6).人体姿势(dense pose)估计：超过3.9万张图像和5.6万个DensePose标注的人物实例。
FiftyOne是一个便于可视化和访问COCO数据资源的开源工具，作为COCO模型分析的评估工具，也可用它来下载COCO数据集，FiftyOne的介绍参考：https://blog.csdn.net/fengbingchun/article/details/121284157 。
COCO-2017数据集总大小约为25.20GB，在FiftyOne中你可以通过”label_types”、 “classes”、“max_samples”参数指定要下载的数据子集；可以通过”image_ids”参数指定要加载的特定图像。详细介绍参考：https://voxel51.com/docs/fiftyone/user_guide/dataset_zoo/datasets.html#dataset-zoo-coco-2017