CrackSeg-MIT-b0-aug

This model is a fine-tuned version of nvidia/mit-b0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0578
• Mean Iou: 0.3169
• Mean Accuracy: 0.6337
• Overall Accuracy: 0.6337
• Accuracy Background: nan
• Accuracy Crack: 0.6337
• Iou Background: 0.0
• Iou Crack: 0.6337

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: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Crack	Iou Background	Iou Crack
0.2102	0.04	100	0.1362	0.1116	0.2232	0.2232	nan	0.2232	0.0	0.2232
0.065	0.08	200	0.1125	0.0153	0.0305	0.0305	nan	0.0305	0.0	0.0305
0.1738	0.12	300	0.1165	0.1976	0.3953	0.3953	nan	0.3953	0.0	0.3953
0.0476	0.17	400	0.1979	0.0120	0.0241	0.0241	nan	0.0241	0.0	0.0241
0.0524	0.21	500	0.1063	0.0533	0.1066	0.1066	nan	0.1066	0.0	0.1066
0.0496	0.25	600	0.1154	0.1646	0.3292	0.3292	nan	0.3292	0.0	0.3292
0.0497	0.29	700	0.0795	0.3184	0.6368	0.6368	nan	0.6368	0.0	0.6368
0.032	0.33	800	0.0905	0.1792	0.3583	0.3583	nan	0.3583	0.0	0.3583
0.1207	0.37	900	0.0738	0.2401	0.4802	0.4802	nan	0.4802	0.0	0.4802
0.0511	0.41	1000	0.0883	0.2591	0.5182	0.5182	nan	0.5182	0.0	0.5182
0.0264	0.46	1100	0.0815	0.1655	0.3309	0.3309	nan	0.3309	0.0	0.3309
0.0719	0.5	1200	0.0772	0.3040	0.6080	0.6080	nan	0.6080	0.0	0.6080
0.042	0.54	1300	0.0707	0.2797	0.5593	0.5593	nan	0.5593	0.0	0.5593
0.167	0.58	1400	0.0685	0.3609	0.7218	0.7218	nan	0.7218	0.0	0.7218
0.0206	0.62	1500	0.0655	0.2469	0.4937	0.4937	nan	0.4937	0.0	0.4937
0.0211	0.66	1600	0.0937	0.3334	0.6668	0.6668	nan	0.6668	0.0	0.6668
0.0659	0.7	1700	0.0750	0.2382	0.4764	0.4764	nan	0.4764	0.0	0.4764
0.0478	0.75	1800	0.0693	0.2944	0.5888	0.5888	nan	0.5888	0.0	0.5888
0.0287	0.79	1900	0.0710	0.2395	0.4790	0.4790	nan	0.4790	0.0	0.4790
0.0359	0.83	2000	0.0580	0.3385	0.6771	0.6771	nan	0.6771	0.0	0.6771
0.0309	0.87	2100	0.0744	0.2153	0.4305	0.4305	nan	0.4305	0.0	0.4305
0.0039	0.91	2200	0.0636	0.2974	0.5947	0.5947	nan	0.5947	0.0	0.5947
0.0152	0.95	2300	0.0635	0.3215	0.6430	0.6430	nan	0.6430	0.0	0.6430
0.0233	0.99	2400	0.0668	0.3039	0.6077	0.6077	nan	0.6077	0.0	0.6077
0.0088	1.04	2500	0.0673	0.3352	0.6704	0.6704	nan	0.6704	0.0	0.6704
0.0756	1.08	2600	0.0599	0.3310	0.6621	0.6621	nan	0.6621	0.0	0.6621
0.0522	1.12	2700	0.0674	0.2943	0.5885	0.5885	nan	0.5885	0.0	0.5885
0.0595	1.16	2800	0.0828	0.2382	0.4763	0.4763	nan	0.4763	0.0	0.4763
0.0135	1.2	2900	0.0574	0.2901	0.5802	0.5802	nan	0.5802	0.0	0.5802
0.0289	1.24	3000	0.0700	0.3186	0.6372	0.6372	nan	0.6372	0.0	0.6372
0.0403	1.28	3100	0.0761	0.3741	0.7483	0.7483	nan	0.7483	0.0	0.7483
0.0131	1.33	3200	0.0600	0.3285	0.6570	0.6570	nan	0.6570	0.0	0.6570
0.0957	1.37	3300	0.0633	0.3400	0.6801	0.6801	nan	0.6801	0.0	0.6801
0.0152	1.41	3400	0.0678	0.3479	0.6958	0.6958	nan	0.6958	0.0	0.6958
0.0235	1.45	3500	0.0636	0.3416	0.6832	0.6832	nan	0.6832	0.0	0.6832
0.0304	1.49	3600	0.0596	0.3606	0.7211	0.7211	nan	0.7211	0.0	0.7211
0.0012	1.53	3700	0.0605	0.2992	0.5983	0.5983	nan	0.5983	0.0	0.5983
0.0435	1.57	3800	0.0563	0.3283	0.6566	0.6566	nan	0.6566	0.0	0.6566
0.05	1.61	3900	0.0601	0.3314	0.6628	0.6628	nan	0.6628	0.0	0.6628
0.063	1.66	4000	0.0617	0.3307	0.6614	0.6614	nan	0.6614	0.0	0.6614
0.0552	1.7	4100	0.0626	0.3580	0.7161	0.7161	nan	0.7161	0.0	0.7161
0.0153	1.74	4200	0.0622	0.2864	0.5728	0.5728	nan	0.5728	0.0	0.5728
0.0446	1.78	4300	0.0612	0.3224	0.6448	0.6448	nan	0.6448	0.0	0.6448
0.0203	1.82	4400	0.0589	0.3167	0.6334	0.6334	nan	0.6334	0.0	0.6334
0.0424	1.86	4500	0.0567	0.3443	0.6887	0.6887	nan	0.6887	0.0	0.6887
0.0103	1.9	4600	0.0591	0.3282	0.6563	0.6563	nan	0.6563	0.0	0.6563
0.0831	1.95	4700	0.0573	0.3224	0.6447	0.6447	nan	0.6447	0.0	0.6447
0.1301	1.99	4800	0.0578	0.3169	0.6337	0.6337	nan	0.6337	0.0	0.6337

Framework versions

• Transformers 4.32.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3