segformer-b0-finetuned-segments-satellite-terrain

This model is a fine-tuned version of nvidia/mit-b0 on the ZeeeWP/terrain_map dataset. It achieves the following results on the evaluation set:

• Loss: 1.6101
• Mean Iou: 0.1907
• Mean Accuracy: 0.3530
• Overall Accuracy: 0.4780
• Accuracy Unlabeled: nan
• Accuracy Sand: 0.5878
• Accuracy Cliff: 0.7663
• Accuracy Bedrock flat: 0.7396
• Accuracy Bedrock lowhill: 0.0135
• Accuracy Bedrock highhill: 0.0
• Accuracy Gravel low hill: 0.3636
• Accuracy Gravel high hill: 0.0
• Iou Unlabeled: 0.0
• Iou Sand: 0.4409
• Iou Cliff: 0.4125
• Iou Bedrock flat: 0.3709
• Iou Bedrock lowhill: 0.0132
• Iou Bedrock highhill: 0.0
• Iou Gravel low hill: 0.2884
• Iou Gravel high hill: 0.0

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Sand	Accuracy Cliff	Accuracy Bedrock flat	Accuracy Bedrock lowhill	Accuracy Bedrock highhill	Accuracy Gravel low hill	Accuracy Gravel high hill	Iou Unlabeled	Iou Sand	Iou Cliff	Iou Bedrock flat	Iou Bedrock lowhill	Iou Bedrock highhill	Iou Gravel low hill	Iou Gravel high hill
1.8047	2.5	20	1.9790	0.1684	0.3173	0.4408	nan	0.4519	0.8376	0.5959	0.0301	0.0	0.3053	0.0	0.0	0.3641	0.3607	0.3496	0.0296	0.0	0.2429	0.0
1.5102	5.0	40	1.7599	0.1694	0.3192	0.4582	nan	0.3901	0.8340	0.6609	0.0192	0.0	0.3305	0.0	0.0	0.3239	0.3799	0.3675	0.0189	0.0	0.2649	0.0
1.4458	7.5	60	1.6088	0.1873	0.3457	0.4639	nan	0.5881	0.7480	0.7632	0.0081	0.0	0.3122	0.0	0.0	0.4691	0.4114	0.3544	0.0080	0.0	0.2555	0.0
1.568	10.0	80	1.6101	0.1907	0.3530	0.4780	nan	0.5878	0.7663	0.7396	0.0135	0.0	0.3636	0.0	0.0	0.4409	0.4125	0.3709	0.0132	0.0	0.2884	0.0

Framework versions

• Transformers 4.44.1
• Pytorch 2.4.1
• Datasets 2.14.6
• Tokenizers 0.19.1