new_ecc_segformer

This model is a fine-tuned version of nvidia/mit-b5 on the rishitunu/ECC_crackdataset_withsplit dataset. It achieves the following results on the evaluation set:

• Loss: 0.0663
• Mean Iou: 0.1943
• Mean Accuracy: 0.3915
• Overall Accuracy: 0.3915
• Accuracy Background: nan
• Accuracy Crack: 0.3915
• Iou Background: 0.0
• Iou Crack: 0.3887

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: 1337
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: polynomial
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Crack	Iou Background	Iou Crack
0.0489	1.0	438	0.0634	0.1464	0.2933	0.2933	nan	0.2933	0.0	0.2929
0.0542	2.0	876	0.0439	0.1956	0.3917	0.3917	nan	0.3917	0.0	0.3912
0.0484	3.0	1314	0.0434	0.1719	0.3551	0.3551	nan	0.3551	0.0	0.3439
0.0539	4.0	1752	0.0447	0.1871	0.3820	0.3820	nan	0.3820	0.0	0.3741
0.0565	5.0	2190	0.0435	0.1888	0.3937	0.3937	nan	0.3937	0.0	0.3777
0.0544	6.0	2628	0.0442	0.1904	0.3930	0.3930	nan	0.3930	0.0	0.3808
0.0421	7.0	3066	0.0449	0.2256	0.4651	0.4651	nan	0.4651	0.0	0.4513
0.0352	8.0	3504	0.0587	0.1569	0.3165	0.3165	nan	0.3165	0.0	0.3138
0.0394	9.0	3942	0.0442	0.1842	0.3710	0.3710	nan	0.3710	0.0	0.3684
0.0445	10.0	4380	0.0609	0.1167	0.4173	0.4173	nan	0.4173	0.0	0.2334
0.0503	11.0	4818	0.0504	0.1702	0.3714	0.3714	nan	0.3714	0.0	0.3403
0.0379	12.0	5256	0.0460	0.1903	0.3869	0.3869	nan	0.3869	0.0	0.3807
0.0405	13.0	5694	0.0452	0.2017	0.4084	0.4084	nan	0.4084	0.0	0.4034
0.0367	14.0	6132	0.0477	0.1995	0.4060	0.4060	nan	0.4060	0.0	0.3990
0.0315	15.0	6570	0.0498	0.2073	0.4208	0.4208	nan	0.4208	0.0	0.4147
0.0244	16.0	7008	0.0486	0.1963	0.4029	0.4029	nan	0.4029	0.0	0.3926
0.031	17.0	7446	0.0568	0.1927	0.3892	0.3892	nan	0.3892	0.0	0.3855
0.0288	18.0	7884	0.0560	0.2033	0.4092	0.4092	nan	0.4092	0.0	0.4067
0.0354	19.0	8322	0.0613	0.2007	0.4056	0.4056	nan	0.4056	0.0	0.4013
0.0315	20.0	8760	0.0605	0.1865	0.3752	0.3752	nan	0.3752	0.0	0.3731
0.0343	21.0	9198	0.0653	0.1991	0.4019	0.4019	nan	0.4019	0.0	0.3981
0.0327	22.0	9636	0.0660	0.1945	0.3924	0.3924	nan	0.3924	0.0	0.3891
0.0252	22.83	10000	0.0663	0.1943	0.3915	0.3915	nan	0.3915	0.0	0.3887

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cpu
• Datasets 2.14.6
• Tokenizers 0.14.1