offroad-mapper-road-segmentation

This model is a fine-tuned version of nvidia/mit-b3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1623
• Mean Iou: 0.4892
• Mean Accuracy: 0.5086
• Bg Iou: 0.9594
• Road Iou: 0.0190

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 6e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: polynomial
• lr_scheduler_warmup_steps: 28
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Bg Iou	Road Iou
3.9395	1.0	29	0.2522	0.4813	0.5013	0.9523	0.0103
1.6425	2.0	58	0.1610	0.4819	0.5010	0.9614	0.0024
1.0853	3.0	87	0.1539	0.4820	0.5011	0.9616	0.0025
1.0067	4.0	116	0.1623	0.4892	0.5086	0.9594	0.0190
0.9354	5.0	145	0.1580	0.4850	0.5040	0.9609	0.0091
0.9454	6.0	174	0.1549	0.4826	0.5017	0.9615	0.0038
0.9599	7.0	203	0.1604	0.4859	0.5050	0.9607	0.0111
0.8990	8.0	232	0.1576	0.4846	0.5036	0.9611	0.0081
0.9328	9.0	261	0.1565	0.4830	0.5020	0.9614	0.0045
0.8854	10.0	290	0.1572	0.4837	0.5027	0.9613	0.0062

Framework versions

• Transformers 5.6.1
• Pytorch 2.6.0+cu126
• Datasets 4.8.4
• Tokenizers 0.22.2