LVIS / README.md
harpreetsahota's picture
Update README.md
6ec5288 verified
metadata
annotations_creators: []
language: en
size_categories:
  - 10K<n<100K
task_categories:
  - object-detection
task_ids: []
pretty_name: LVIS-35k
tags:
  - fiftyone
  - image
  - object-detection
dataset_summary: >




  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 35000
  samples.


  ## Installation


  If you haven't already, install FiftyOne:


  ```bash

  pip install -U fiftyone

  ```


  ## Usage


  ```python

  import fiftyone as fo

  import fiftyone.utils.huggingface as fouh


  # Load the dataset

  # Note: other available arguments include 'max_samples', etc

  dataset = fouh.load_from_hub("Voxel51/LVIS")


  # Launch the App

  session = fo.launch_app(dataset)

  ```

Dataset Card for LVIS-35k

image

This is a FiftyOne dataset with 35000 samples.

NOTE: This is only a 35k sample subset of the full dataset. The notebook recipe for creating this, and the full, dataset can be found here

Installation

If you haven't already, install FiftyOne:

pip install -U fiftyone

Usage

import fiftyone as fo
import fiftyone.utils.huggingface as fouh

# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = fouh.load_from_hub("Voxel51/LVIS")

# Launch the App
session = fo.launch_app(dataset)

Dataset Details

Dataset Description

LVIS (pronounced 'el-vis') is a dataset for large vocabulary instance segmentation, introduced by researchers from Facebook AI.

  • It contains annotations for over 1000 object categories across 164k images. The full dataset is planned to have ~2 million high-quality instance segmentation masks.

  • The categories in LVIS follow a natural long-tail distribution, with a few common categories and many rare ones with few training examples. This long tail poses a challenge for current state-of-the-art object detection methods which struggle with low-sample categories.

  • The vocabulary was constructed iteratively, starting from 8.8k concrete noun synsets in WordNet and filtering down to the final set[4].

  • LVIS can be used for instance segmentation, semantic segmentation, and object detection tasks. The dataset aims to focus the research community on the open challenge of long-tail object recognition.

In summary, LVIS is a large-scale, high-quality dataset that targets the difficult problem of learning segmentation models for various object categories, including many rare ones. It is freely available for research use.

  • Curated by: Agrim Gupta, Piotr Dollár, Ross Girshick
  • Funded by: Facebook AI Research (FAIR)
  • Shared by: Harpreet Sahota, Hacker-in-Residence at Voxel51
  • Language(s) (NLP): en
  • License: Custom License

Dataset Sources [optional]

Citation

BibTeX:

@inproceedings{gupta2019lvis,
  title={{LVIS}: A Dataset for Large Vocabulary Instance Segmentation},
  author={Gupta, Agrim and Dollar, Piotr and Girshick, Ross},
  booktitle={Proceedings of the {IEEE} Conference on Computer Vision and Pattern Recognition},
  year={2019}
}