---
license: mit
datasets:
- CRASAR/CRASAR-U-DROIDs
---

This repository contains a Road Damage Assessment model trained on the [CRASAR-U-DROIDs Road Damage Assessment dataset](https://huggingface.co/datasets/CRASAR/CRASAR-U-DROIDs). The following files are provided...
1) The [model file](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/blob/main/RDA_SegmenterVit_NoPreTrain_CCE_simple-epoch%3D04-step%3D1250-val_macro_iou%3D0.15187.ckpt).
2) The set of predicted labels generated by the model on the CRASAR-U-DROIDs dataset for both [adjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/resolve/main/CRASAR-U-DROIDs_RDA_preds_adjusted.json?download=true) and [unadjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/resolve/main/CRASAR-U-DROIDs_RDA_preds_unadjusted.json?download=true) formulations that were used for evaluation.
3) Metrics files generated from test set evaluation on the CRASAR-U-DROIDs dataset for both [adjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/blob/main/CRASAR-U-DROIDs_RDA_metrics_adjusted.json) and [unadjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/blob/main/CRASAR-U-DROIDs_RDA_metrics_unadjusted.json) predicted labels.
4) The full set of hyperparameters used to train and evaluate the model for both [adjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/blob/main/RDA_ScaleMAE_NoPretrain_Segmenter_simple_adjusted.yaml) and [unadjusted](https://huggingface.co/CRASAR/CRASAR-U-DROIDs-RDA_AAAI26_Simple_NoPretrained_ScaleMAE_Segmenter/blob/main/RDA_ScaleMAE_NoPretrain_Segmenter_simple_unadjusted.yaml) formulations.