SegFormer_Mixed_Set2_Grayscale_mit-b5

This model is a fine-tuned version of nvidia/mit-b5 on Hasano20/Mixed_Set2_Grayscale dataset. It achieves the following results on the evaluation set:

• Train-Loss: 0.0081
• Loss: 0.0138
• Mean Iou: 0.9805
• Mean Accuracy: 0.9909
• Overall Accuracy: 0.9952
• Accuracy Background: 0.9959
• Accuracy Melt: 0.9801
• Accuracy Substrate: 0.9967
• Iou Background: 0.9926
• Iou Melt: 0.9554
• Iou Substrate: 0.9934

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Melt	Accuracy Substrate	Iou Background	Iou Melt	Iou Substrate
0.1464	0.7042	50	0.1373	0.8536	0.9233	0.9584	0.9574	0.8385	0.9741	0.9444	0.6783	0.9381
0.0813	1.4085	100	0.0616	0.9196	0.9481	0.9808	0.9873	0.8668	0.9902	0.9748	0.8112	0.9728
0.0415	2.1127	150	0.0608	0.9074	0.9281	0.9802	0.9931	0.7978	0.9934	0.9769	0.7718	0.9734
0.0486	2.8169	200	0.0344	0.9475	0.9675	0.9876	0.9859	0.9190	0.9976	0.9836	0.8764	0.9824
0.028	3.5211	250	0.0226	0.9663	0.9845	0.9922	0.9929	0.9658	0.9949	0.9894	0.9205	0.9892
0.0309	4.2254	300	0.0214	0.9686	0.9839	0.9924	0.9931	0.9630	0.9956	0.9894	0.9273	0.9891
0.0136	4.9296	350	0.0248	0.9637	0.9828	0.9913	0.9901	0.9623	0.9959	0.9885	0.9151	0.9873
0.021	5.6338	400	0.0182	0.9717	0.9881	0.9933	0.9942	0.9752	0.9949	0.9908	0.9338	0.9906
0.0178	6.3380	450	0.0163	0.9747	0.9907	0.9940	0.9945	0.9826	0.9950	0.9913	0.9409	0.9918
0.0211	7.0423	500	0.0167	0.9746	0.9877	0.9939	0.9949	0.9725	0.9958	0.9911	0.9414	0.9913
0.0161	7.7465	550	0.0162	0.9751	0.9883	0.9939	0.9936	0.9747	0.9966	0.9910	0.9429	0.9914
0.0128	8.4507	600	0.0145	0.9769	0.9903	0.9944	0.9940	0.9805	0.9965	0.9916	0.9468	0.9924
0.0132	9.1549	650	0.0150	0.9780	0.9891	0.9946	0.9946	0.9757	0.9970	0.9918	0.9498	0.9923
0.0118	9.8592	700	0.0144	0.9775	0.9907	0.9946	0.9938	0.9815	0.9968	0.9915	0.9483	0.9927
0.0088	10.5634	750	0.0136	0.9792	0.9907	0.9949	0.9952	0.9804	0.9965	0.9922	0.9524	0.9930
0.0085	11.2676	800	0.0140	0.9789	0.9904	0.9948	0.9947	0.9797	0.9968	0.9921	0.9517	0.9929
0.0109	11.9718	850	0.0142	0.9782	0.9919	0.9948	0.9950	0.9849	0.9959	0.9921	0.9497	0.9929
0.009	12.6761	900	0.0134	0.9799	0.9908	0.9951	0.9951	0.9804	0.9969	0.9923	0.9542	0.9933
0.0105	13.3803	950	0.0135	0.9797	0.9912	0.9951	0.9953	0.9817	0.9966	0.9923	0.9536	0.9933
0.0094	14.0845	1000	0.0142	0.9786	0.9868	0.9948	0.9953	0.9673	0.9979	0.9923	0.9509	0.9927
0.0089	14.7887	1050	0.0136	0.9799	0.9907	0.9951	0.9955	0.9800	0.9967	0.9924	0.9541	0.9933
0.0118	15.4930	1100	0.0140	0.9794	0.9897	0.9950	0.9962	0.9763	0.9965	0.9924	0.9528	0.9932
0.0101	16.1972	1150	0.0142	0.9792	0.9914	0.9950	0.9950	0.9828	0.9965	0.9922	0.9521	0.9933
0.0081	16.9014	1200	0.0182	0.9748	0.9844	0.9942	0.9961	0.9601	0.9970	0.9923	0.9405	0.9915
0.0111	17.6056	1250	0.0154	0.9772	0.9913	0.9945	0.9942	0.9837	0.9961	0.9918	0.9471	0.9925
0.0078	18.3099	1300	0.0136	0.9800	0.9905	0.9951	0.9958	0.9791	0.9966	0.9925	0.9544	0.9933
0.0059	19.0141	1350	0.0139	0.9802	0.9915	0.9952	0.9953	0.9824	0.9967	0.9926	0.9545	0.9934
0.0081	19.7183	1400	0.0138	0.9805	0.9909	0.9952	0.9959	0.9801	0.9967	0.9926	0.9554	0.9934

Framework versions

• Transformers 4.41.2
• Pytorch 2.0.1+cu117
• Datasets 2.19.2
• Tokenizers 0.19.1