数据集:

winvoker/lvis

中文

LVIS

Dataset Summary

This dataset is the implementation of LVIS dataset into Hugging Face datasets. Please visit the original website for more information.

Loading

This code returns train, validation and test generators.

from datasets import load_dataset

dataset = load_dataset("winvoker/lvis")

Objects is a dictionary which contains annotation information like bbox, class.

DatasetDict({
    train: Dataset({
        features: ['id', 'image', 'height', 'width', 'objects'],
        num_rows: 100170
    })
    validation: Dataset({
        features: ['id', 'image', 'height', 'width', 'objects'],
        num_rows: 4809
    })
    test: Dataset({
        features: ['id', 'image', 'height', 'width', 'objects'],
        num_rows: 19822
    })
})

Access Generators

train = dataset["train"]
validation = dataset["validation"]
test = dataset["test"] 

An example row is as follows.

{  'id': 0, 
   'image': '000000437561.jpg', 
   'height': 480, 
   'width': 640, 
   'objects': {
               'bboxes': [[[392, 271, 14, 3]], 
               'classes': [117], 
               'segmentation': [[376, 272, 375, 270, 372, 269, 371, 269, 373, 269, 373]]
               }
 }