https://github.com/quiltdata/open-fruit
Code and data for the Open Fruit data package
https://github.com/quiltdata/open-fruit
Last synced: 29 days ago
JSON representation
Code and data for the Open Fruit data package
- Host: GitHub
- URL: https://github.com/quiltdata/open-fruit
- Owner: quiltdata
- Created: 2019-04-16T17:48:12.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2019-04-17T19:55:07.000Z (over 7 years ago)
- Last Synced: 2025-10-25T04:23:58.399Z (10 months ago)
- Language: Jupyter Notebook
- Size: 1.03 MB
- Stars: 3
- Watchers: 5
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# open fruit
This data package defines the "Open Fruit" dataset. Open Fruit is a collection of raw and cropped images and associated metadata of fruits taken from the [Google Open Images V4 dataset](https://github.com/openimages/dataset), which was generated by scraping and classifying permissively-licensed images on [flickr](https://www.flickr.com/). You can view this data package on [GitHub](https://github.com/quiltdata/open-fruit) (code only) or on [Quilt T4](https://alpha.quiltdata.com/b/quilt-example/tree/quilt/open_fruit/) (code + data).
To learn more about Open Images check out ["How to classify photos in 600 classes using nine million open images"](https://blog.quiltdata.com/how-to-classify-photos-in-600-classes-using-nine-million-open-images-3cdb989ad1c2).
This data package is used for a pair of articles:
* ["Boost your CNN image classifier performance with progressive resizing in Keras"](https://blog.quiltdata.com/boost-your-cnn-image-classifier-performance-with-progressive-resizing-in-keras-c215127bf55b)
* ["Building a fully reproducible machine learning pipeline with Comet.ml and Quilt"](https://medium.com/@ceceliashao/aa9c7bf85e72)
See also the [`ResidentMario/progressive-resizing`](https://github.com/ResidentMario/progressive-resizing) and [`comet-ml/keras-fruit-classifier`](https://github.com/comet-ml/keras-fruit-classifer) repositories, which contain the model builds that use this package.
To rebuild this dataset from scratch:
1. Run `notebooks/initial-exploration.ipynb`, to download the image metadata.
2. Download the [`openimager` script](https://github.com/quiltdata/open-images/blob/master/src/openimager/openimager.py) and run `python -c "import openimager; openimager.download(['Apple', 'Banana', 'Cantaloupe', 'Common fig', 'Grape', 'Grapefruit', 'Lemon', 'Mango', 'Orange', 'Peach', 'Pear', 'Pineapple', 'Pomegranate', 'Watermelon', 'Strawberry', 'Tomato'])"` (warning: this will take some time).
3. Run `notebooks/build-dataset.ipynb` to crop the images and split them by class (this will also take some time).