FINAL_ecc_segformer

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

• Loss: 0.0749
• Mean Iou: 0.1968
• Mean Accuracy: 0.3939
• Overall Accuracy: 0.3939
• Accuracy Background: nan
• Accuracy Crack: 0.3939
• Iou Background: 0.0
• Iou Crack: 0.3936

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
0.0534	1.0	548	0.0614	0.1368	0.2750	0.2750	nan	0.2750	0.0	0.2736
0.058	2.0	1096	0.1018	0.2093	0.4238	0.4238	nan	0.4238	0.0	0.4186
0.0482	3.0	1644	0.0508	0.1791	0.4315	0.4315	nan	0.4315	0.0	0.3582
0.0338	4.0	2192	0.0569	0.1849	0.3716	0.3716	nan	0.3716	0.0	0.3698
0.0395	5.0	2740	0.0597	0.1745	0.3506	0.3506	nan	0.3506	0.0	0.3490
0.0372	6.0	3288	0.0509	0.2298	0.4635	0.4635	nan	0.4635	0.0	0.4597
0.0402	7.0	3836	0.0620	0.1751	0.3507	0.3507	nan	0.3507	0.0	0.3503
0.038	8.0	4384	0.0681	0.1905	0.3815	0.3815	nan	0.3815	0.0	0.3810
0.0393	9.0	4932	0.0685	0.2213	0.4433	0.4433	nan	0.4433	0.0	0.4425
0.0376	10.0	5480	0.0590	0.1962	0.3929	0.3929	nan	0.3929	0.0	0.3924
0.0381	11.0	6028	0.0626	0.1891	0.3801	0.3801	nan	0.3801	0.0	0.3783
0.034	12.0	6576	0.0623	0.2061	0.4162	0.4162	nan	0.4162	0.0	0.4122
0.0301	13.0	7124	0.0831	0.1832	0.3669	0.3669	nan	0.3669	0.0	0.3664
0.034	14.0	7672	0.0636	0.2059	0.4119	0.4119	nan	0.4119	0.0	0.4118
0.0303	15.0	8220	0.0705	0.1931	0.3864	0.3864	nan	0.3864	0.0	0.3862
0.0338	16.0	8768	0.0685	0.2101	0.4206	0.4206	nan	0.4206	0.0	0.4202
0.0229	17.0	9316	0.0706	0.2099	0.4204	0.4204	nan	0.4204	0.0	0.4197
0.0337	18.0	9864	0.0742	0.1982	0.3968	0.3968	nan	0.3968	0.0	0.3965
0.0257	18.25	10000	0.0749	0.1968	0.3939	0.3939	nan	0.3939	0.0	0.3936

Framework versions

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