sarapohland/park-navigation

This dataset was obtained from a quadraped traversing a wooded park. The dataset contains 13,839 RGB images (sized 240x320) taken from different regions within the park environment that appear visually different. All images are labeled as corresponding to one of eight regions:

1. Grassy region at center of park (3,264 images)
2. Grassy region at left side of park (1,566 images)
3. Grassy region at right side of park (968 images)
4. Grassy region near building (1,598 images)
5. Grassy region near parking lot (1,366 images)
6. Grassy region near pavilion (935 images)
7. Wooded region of park (2,192 images)
8. Region very close to a tree (1,950 images)

This labeled dataset has been randomly partitioned into a training set of 9,964 images, a validation set of 2,491 images, and a test set of 1,384 images.

This dataset also contains a set of out-of-distribution (OOD) data points. The in-distribution data contains images from grassy and forested region of the park, in which images contain grass, trees, the sky, etc. OOD data contains images from around the park's pavilion, in which images additionally contain various human-made structures (conrete, trash cans, park benches, etc.). There are 298 RGB images (sized 240x320) in the OOD set.

The compressed NumPy file contains the following elements: train_data, train_labels, val_data, val_labels, test_data, test_labels, ood_data, and ood_labels. Sample code for interacting with this dataset is available in our GitHub repository. In particular, src/datasets/custom_dataset.py demonstrates how to set up a PyTorch Dataset and src/datasets/setup_dataloader.py demonstrates how to set up a PyTorch Dataloader.

Each pixel in each of the OOD images is either familiar or unfamiliar to the perception model, depending on whether it corresponds to structures present in the training set or not. These familiarity labels are stored in the ood_labels pickle file. To adjust these labels, you can use the src/datasets/create_labels.py script available in our GitHub repository.