LVIS_v1_dot_0

displayName: LVIS v1.0
labelTypes:
- Box2d
license:
- CC BY 4.0
mediaTypes:
- Image
paperUrl: https://arxiv.org/pdf/1908.03195.pdf
publishDate: "2019"
publishUrl: https://www.lvisdataset.org/
publisher:
- Facebook AI Research
tags:
- Image
taskTypes:
- Instance Segmentation
- 2D-ObjectDetection

---
# 数据集介绍
## 简介
将研究社区的注意力集中在开放挑战上的数据集推动了对象检测的进展。这个过程使我们从简单的图像到复杂的场景，从边界框到分割掩码。在这项工作中，我们介绍了 LVIS（发音为 'el-vis'）：一个用于大型词汇实例分割的新数据集。我们计划为 164k 图像中的 1000 多个入门级对象类别收集约 200 万个高质量实例分割掩码。由于自然图像中类别的 Zipfian 分布，LVIS 自然有一个长尾类别，训练样本很少。鉴于用于对象检测的最先进的深度学习方法在低样本情况下表现不佳，我们相信我们的数据集提出了一个重要且令人兴奋的新科学挑战。所有LVIS数据集图像均来自COCO数据集，图像使用条款可参考COCO条款

## Download dataset
:modelscope-code[]{type="git"}

displayName: LVIS v1.0
labelTypes:
- Box2d
license:
- CC BY 4.0
mediaTypes:
- Image
paperUrl: https://arxiv.org/pdf/1908.03195.pdf
publishDate: "2019"
publishUrl: https://www.lvisdataset.org/
publisher:
- Facebook AI Research
tags:
- Image
taskTypes:
- Instance Segmentation
- 2D-ObjectDetection

---
# Dataset Overview
## Introduction
Datasets that center the research community's attention on open challenges have driven the advancement of object detection. This process has propelled the field from simple images to complex scenes, and from bounding boxes to segmentation masks. In this work, we introduce LVIS (pronounced 'el-vis'): a novel dataset for Large Vocabulary Instance Segmentation. We plan to collect approximately 2 million high-quality instance segmentation masks for over 1000 entry-level object categories across 164k images. Owing to the Zipfian distribution of object categories in natural images, LVIS naturally has a long-tailed category distribution with limited training samples. Given that state-of-the-art deep learning methods for object detection perform poorly in low-sample scenarios, we believe that our dataset presents an important and exciting new scientific challenge. All images in the LVIS dataset are sourced from the COCO dataset, and the terms of use for these images follow the COCO terms of use.

## Download Dataset
:modelscope-code[]{type="git"}