test-carbonate-segmentation2

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

• Loss: 0.3642
• Mean Iou: 0.2180
• Mean Accuracy: 0.3344
• Overall Accuracy: 0.6454
• Accuracy Micrite: nan
• Accuracy Cement: nan
• Accuracy Peloid/pellet/ooid: nan
• Accuracy Biotic: 0.6660
• Accuracy Scale bar: 0.0028
• Iou Micrite: 0.0
• Iou Cement: nan
• Iou Peloid/pellet/ooid: nan
• Iou Biotic: 0.6511
• Iou Scale bar: 0.0028

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Micrite	Accuracy Cement	Accuracy Peloid/pellet/ooid	Accuracy Biotic	Accuracy Scale bar	Iou Micrite	Iou Cement	Iou Peloid/pellet/ooid	Iou Biotic	Iou Scale bar
1.2003	2.22	20	1.5260	0.4834	0.7334	0.9835	nan	nan	nan	1.0	0.4669	0.0	nan	nan	0.9834	0.4669
1.2006	4.44	40	0.8923	0.6346	0.9521	0.9498	nan	nan	nan	0.9497	0.9545	0.0	nan	nan	0.9494	0.9545
1.4233	6.67	60	1.0240	0.4417	0.6716	0.9793	nan	nan	nan	0.9995	0.3438	0.0	nan	nan	0.9814	0.3438
1.1735	8.89	80	0.7964	0.5230	0.7890	0.9437	nan	nan	nan	0.9539	0.6241	0.0	nan	nan	0.9449	0.6241
1.0242	11.11	100	0.8747	0.5322	0.8038	0.9849	nan	nan	nan	0.9969	0.6108	0.0	nan	nan	0.9859	0.6108
0.9161	13.33	120	0.9217	0.5133	0.7767	0.9831	nan	nan	nan	0.9967	0.5568	0.0	nan	nan	0.9832	0.5568
0.8102	15.56	140	0.7069	0.5923	0.8907	0.9490	nan	nan	nan	0.9529	0.8286	0.0	nan	nan	0.9484	0.8286
0.5436	17.78	160	0.5149	0.3206	0.4929	0.8806	nan	nan	nan	0.9062	0.0795	0.0	nan	nan	0.8823	0.0795
0.8517	20.0	180	0.5646	0.3748	0.5719	0.9200	nan	nan	nan	0.9430	0.2008	0.0	nan	nan	0.9236	0.2008
0.4532	22.22	200	0.6128	0.3133	0.4837	0.9188	nan	nan	nan	0.9475	0.0199	0.0	nan	nan	0.9201	0.0199
1.3133	24.44	220	0.3006	0.2391	0.3645	0.7064	nan	nan	nan	0.7290	0.0	0.0	nan	nan	0.7172	0.0
0.4636	26.67	240	0.3260	0.1903	0.2901	0.5259	nan	nan	nan	0.5414	0.0388	0.0	nan	nan	0.5320	0.0388
0.9843	28.89	260	0.3663	0.2741	0.4182	0.6986	nan	nan	nan	0.7171	0.1193	0.0	nan	nan	0.7031	0.1193
0.7617	31.11	280	0.3338	0.2357	0.3627	0.7030	nan	nan	nan	0.7255	0.0	0.0	nan	nan	0.7072	0.0
1.283	33.33	300	0.3395	0.2723	0.4176	0.7232	nan	nan	nan	0.7434	0.0919	0.0	nan	nan	0.7250	0.0919
0.6578	35.56	320	0.3382	0.2069	0.3170	0.6143	nan	nan	nan	0.6339	0.0	0.0	nan	nan	0.6207	0.0
0.2129	37.78	340	0.3436	0.2288	0.3525	0.6831	nan	nan	nan	0.7049	0.0	0.0	nan	nan	0.6863	0.0
0.7001	40.0	360	0.2998	0.2001	0.3069	0.5771	nan	nan	nan	0.5950	0.0189	0.0	nan	nan	0.5813	0.0189
0.3866	42.22	380	0.3162	0.1840	0.2819	0.5464	nan	nan	nan	0.5639	0.0	0.0	nan	nan	0.5521	0.0
1.2623	44.44	400	0.3431	0.2125	0.3254	0.6172	nan	nan	nan	0.6365	0.0142	0.0	nan	nan	0.6234	0.0142
0.6115	46.67	420	0.2987	0.2020	0.3095	0.5998	nan	nan	nan	0.6190	0.0	0.0	nan	nan	0.6060	0.0
0.5802	48.89	440	0.3642	0.2180	0.3344	0.6454	nan	nan	nan	0.6660	0.0028	0.0	nan	nan	0.6511	0.0028

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.1+cu118
• Tokenizers 0.13.3