mit-b0-building-damage-lora

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

• Loss: 0.0661
• Mean Iou: 0.3623
• Mean Accuracy: 0.7245
• Overall Accuracy: 0.7245
• Accuracy Building: 0.7245
• Iou Building: 0.7245

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Building	Iou Building
0.0618	1.0	700	0.1463	0.4063	0.8125	0.8125	0.8125	0.8125
0.0813	2.0	1400	0.0861	0.3950	0.7900	0.7900	0.7900	0.7900
0.0715	3.0	2100	0.0856	0.3844	0.7689	0.7689	0.7689	0.7689
0.076	4.0	2800	0.1296	0.4161	0.8322	0.8322	0.8322	0.8322
0.0587	5.0	3500	0.0702	0.3078	0.6156	0.6156	0.6156	0.6156
0.0662	6.0	4200	0.0708	0.3613	0.7226	0.7226	0.7226	0.7226
0.059	7.0	4900	0.1063	0.4125	0.8249	0.8249	0.8249	0.8249
0.0532	8.0	5600	0.0693	0.3547	0.7094	0.7094	0.7094	0.7094
0.066	9.0	6300	0.0754	0.3932	0.7863	0.7863	0.7863	0.7863
0.0628	10.0	7000	0.0692	0.3874	0.7747	0.7747	0.7747	0.7747
0.0805	11.0	7700	0.0701	0.3896	0.7793	0.7793	0.7793	0.7793
0.0595	12.0	8400	0.0663	0.3774	0.7549	0.7549	0.7549	0.7549
0.0705	13.0	9100	0.0653	0.3717	0.7433	0.7433	0.7433	0.7433
0.071	14.0	9800	0.0651	0.3731	0.7461	0.7461	0.7461	0.7461
0.0656	15.0	10500	0.0648	0.3613	0.7227	0.7227	0.7227	0.7227

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.13.3