Install the ECOLE dataset package.
pip3 install git+https://github.com/xingyaoww/ecole-dataset.git
For example, you can load all the evaluation datasets by running the following code:
import ecole_dataset
test_datasets = ecole_dataset.registry.load_datasets(
split=ecole_dataset.SplitType.TEST
)
Please check scripts/example_usecase.ipynb
for more examples.
NOTE: This package currently only supports loading dataset metadata (i.e., the test_datasets
returned by .load_datasets
does not contains pre-loaded images, but only image filepaths). We are actively working on allowing images to be also automatically loaded into test_datasets
with a single line of code.
Please refer to docs/IMAGE_DATA.md to learn about how to download images and configure so that ecole-dataset
can automatically loads images with one line of code.
Datasets | Concept Types | Format | HF Dataset Link | Official Repo/Homepage |
---|---|---|---|---|
OmniLabel | object | image | https://huggingface.co/datasets/xingyaoww/omnilabel | https://www.omnilabel.org/dataset/download |
VAW | attribute, object | image | https://huggingface.co/datasets/mikewang/vaw | https://github.com/adobe-research/vaw_dataset#dataset-setup |
PAD-v2 | affordance | image | https://huggingface.co/datasets/mikewang/padv2 | https://github.com/lhc1224/OSAD_Net |
ImSitu | activity | image | https://huggingface.co/datasets/mikewang/imsitu | https://github.com/my89/imSitu |
Kinetics-400 | activity | video | https://huggingface.co/datasets/AlexFierro9/Kinetics400 | https://www.deepmind.com/open-source/kinetics |
AwA2 | attribute (Animal) | image | https://huggingface.co/datasets/mikewang/AwA2 | https://cvml.ista.ac.at/AwA2/ |
EuroSAT | object (Satellite Imagery) | image | https://huggingface.co/datasets/mikewang/EuroSAT | https://github.com/phelber/EuroSAT |
Datasets | Concept Types | Format | HF Dataset Link | Official Repo/Homepage |
---|---|---|---|---|
Visual Genome | object, attribute, relationship | image | https://huggingface.co/datasets/visual_genome | https://homes.cs.washington.edu/~ranjay/visualgenome/index.html |