Modelpublic

roboticslaburjc-org/cross-dataset-visual-offroad-segmentation

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face

1likes

Model Card

Cross-Dataset Evaluation of Visual Semantic Segmentation Models for Off-Road Autonomous Driving

David Pascual-Hernández, Sergio Paniego, Roberto Calvo-Palomino, Inmaculada Mora-Jiménez, Jose María Cañas-Plaza

We provide weights for the models studied in Cross-Dataset Evaluation of Visual Semantic Segmentation Models for Off-Road Autonomous Driving. Directories nomenclature is defined as:

<dataset>-<ontology>-<data_augmentation>```


- **Dataset**. Training dataset, either RELLIS-3D, GOOSE or a combination of both.
- **Ontology**. Ontology used during training. It can either be the original dataset ontology or our unified coarser ontology.
- **Data augmentation**. Data augmentation strategy, either *light* (used for inner-dataset experiments) or *strong* (used for cross-dataset experiments).

All models have been fine-tuned using the pre-trained weights provided by [mmsegmentation](https://github.com/open-mmlab/mmsegmentation/tree/main) for the Cityscapes dataset.