Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/lorenzoh/visiondatasets.jl
Download and load common computer vision datasets for different tasks
https://github.com/lorenzoh/visiondatasets.jl
Last synced: 11 days ago
JSON representation
Download and load common computer vision datasets for different tasks
- Host: GitHub
- URL: https://github.com/lorenzoh/visiondatasets.jl
- Owner: lorenzoh
- Created: 2019-12-28T15:52:04.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2020-09-12T09:36:12.000Z (over 4 years ago)
- Last Synced: 2024-11-11T16:44:14.277Z (2 months ago)
- Language: Julia
- Size: 67.4 KB
- Stars: 1
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# VisionDatasets.jl
Download and load common computer vision datasets for different tasks.
Datasets are grouped by task, so far including:
- classification
This package uses the dataset interface of [MLDataPattern.jl](https://github.com/JuliaML/MLDataPattern.jl), meaning that every dataset defines:
- `LearnBase.getobs(ds, idx|idxs)` to load one or more samples
- `LearnBase.nobs(ds)` to get the number of samples in a dataset
Samples are represented as `Dict`s, and not as tuples of `(x, y)` because there is often more information associated with each sample, like the dataset split is part of.
The `Dict`s keys differ for every task, see below for more info.
## Install
```julia
]add https://www.github.com/lorenzoh/VisionDatasets.jl
```
## Usage
```julia
using VisionDatasets
using LearnBase: getobs, nobs
# load dataset files and labels
> dataset = ImageWoof.ImageWoof2_160()
ClassificationDataset(...)
# grab a sample
> getobs(dataset, 1)
Dict(:image => ..., :label => ..., ...)
```
## Datasets
### Classification
A sample of a `ClassificationDataset` is a `Dict` with the following keys:
- `:image`
- `:label`: the sample's label id, from 1 to number of classes; check `dataset.names` to access to corresponding label names
#### Available classification datasets
- [ImageWoof](https://github.com/fastai/imagenette)
- ImageWoof2 (160px): `ImageWoof.ImageWoof2_160`
- ImageWoof2 (320px): `ImageWoof.ImageWoof2_320`
- [ImageNette](https://github.com/fastai/imagenette)
- ImageNette2 (160 px): `ImageNette.ImageNette2_160`
- ImageNette2 (320 px): `ImageNette.ImageNette2_320`
### Pose estimation
A sample of a `ClassificationDataset` is a `Dict` with the following keys:
- `:image`
- `:poses`
- `:config`: The `PoseEstimation.PoseConfig` of the dataset
#### Available pose estimation datasets
- [MPII]: `MPII.mpii(imagespath)`
- [MPII]: `COCO.coco_keypoints(imagespath)`