setup_datasets.md

December 28, 2023 · View on GitHub

Setup Evaluation Datasets

Below are the datasets used in our evaluations, along with links for downloading:

Alternatively, for ease of use and standardization, you may prefer to download a preprocessed version of the datasets from Synboost, which includes RoadAnomaly, FS Static, and FS Lost&Found:

Preprocessed Datasets

After downloading the dataset(s), you will need to configure the data directory in data_set.py to align with the structure of the downloaded files. The expected data structure for these datasets in our code is as follows:

fishyscapes/
├── LostAndFound
│   ├── labels
│   ├── original
│   ├── semantic
└── Static
    ├── labels
    ├── original
    └── semantic

road_anomaly/
├── labels
├── original
├── semantic
└── ...

To quantitatively investigate the impact of domain shift on existing OOD detection methods, we modify the original public Fishyscapes Static dataset with post-load transformations, including random smog, color shifting, and Gaussian blur. You can activate these conditions by setting '--dataset FS_Static_C' in the command, as detailed in README.md.