segformer-b0-harimehta-foc-feb21

This model is a fine-tuned version of nvidia/mit-b0 on the hari000/foc-segmentation dataset. It achieves the following results on the evaluation set:

• Loss: 0.0202
• Mean Iou: 0.2651
• Mean Accuracy: 0.5301
• Overall Accuracy: 0.5301
• Accuracy Cask: nan
• Accuracy Foc: 0.5301
• Iou Cask: 0.0
• Iou Foc: 0.5301

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 45

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Cask	Accuracy Foc	Iou Cask	Iou Foc
0.5516	0.625	20	0.5585	0.0794	0.1588	0.1588	nan	0.1588	0.0	0.1588
0.4061	1.25	40	0.3763	0.0441	0.0882	0.0882	nan	0.0882	0.0	0.0882
0.3363	1.875	60	0.2669	0.0023	0.0046	0.0046	nan	0.0046	0.0	0.0046
0.2396	2.5	80	0.2026	0.0010	0.0021	0.0021	nan	0.0021	0.0	0.0021
0.1909	3.125	100	0.1579	0.0002	0.0004	0.0004	nan	0.0004	0.0	0.0004
0.1587	3.75	120	0.1315	0.0002	0.0003	0.0003	nan	0.0003	0.0	0.0003
0.136	4.375	140	0.1181	0.0001	0.0001	0.0001	nan	0.0001	0.0	0.0001
0.1239	5.0	160	0.0908	0.0	0.0	0.0	nan	0.0	0.0	0.0
0.0973	5.625	180	0.0839	0.0352	0.0704	0.0704	nan	0.0704	0.0	0.0704
0.093	6.25	200	0.0745	0.0366	0.0732	0.0732	nan	0.0732	0.0	0.0732
0.0757	6.875	220	0.0672	0.0013	0.0025	0.0025	nan	0.0025	0.0	0.0025
0.0634	7.5	240	0.0628	0.0281	0.0562	0.0562	nan	0.0562	0.0	0.0562
0.0762	8.125	260	0.0573	0.0337	0.0673	0.0673	nan	0.0673	0.0	0.0673
0.0512	8.75	280	0.0545	0.0317	0.0634	0.0634	nan	0.0634	0.0	0.0634
0.0542	9.375	300	0.0535	0.0727	0.1454	0.1454	nan	0.1454	0.0	0.1454
0.0489	10.0	320	0.0496	0.0583	0.1166	0.1166	nan	0.1166	0.0	0.1166
0.0361	10.625	340	0.0441	0.0586	0.1173	0.1173	nan	0.1173	0.0	0.1173
0.0366	11.25	360	0.0417	0.0551	0.1102	0.1102	nan	0.1102	0.0	0.1102
0.05	11.875	380	0.0416	0.0692	0.1383	0.1383	nan	0.1383	0.0	0.1383
0.04	12.5	400	0.0393	0.0748	0.1495	0.1495	nan	0.1495	0.0	0.1495
0.0321	13.125	420	0.0369	0.1059	0.2118	0.2118	nan	0.2118	0.0	0.2118
0.0355	13.75	440	0.0365	0.0923	0.1846	0.1846	nan	0.1846	0.0	0.1846
0.0296	14.375	460	0.0374	0.1721	0.3443	0.3443	nan	0.3443	0.0	0.3443
0.0395	15.0	480	0.0338	0.0990	0.1980	0.1980	nan	0.1980	0.0	0.1980
0.0294	15.625	500	0.0322	0.1325	0.2650	0.2650	nan	0.2650	0.0	0.2650
0.034	16.25	520	0.0327	0.1611	0.3223	0.3223	nan	0.3223	0.0	0.3223
0.0339	16.875	540	0.0311	0.1840	0.3680	0.3680	nan	0.3680	0.0	0.3680
0.0365	17.5	560	0.0310	0.2204	0.4408	0.4408	nan	0.4408	0.0	0.4408
0.0185	18.125	580	0.0289	0.1810	0.3620	0.3620	nan	0.3620	0.0	0.3620
0.0382	18.75	600	0.0296	0.1902	0.3804	0.3804	nan	0.3804	0.0	0.3804
0.0213	19.375	620	0.0285	0.1498	0.2997	0.2997	nan	0.2997	0.0	0.2997
0.0225	20.0	640	0.0280	0.1862	0.3724	0.3724	nan	0.3724	0.0	0.3724
0.0212	20.625	660	0.0271	0.1821	0.3642	0.3642	nan	0.3642	0.0	0.3642
0.0292	21.25	680	0.0270	0.1965	0.3929	0.3929	nan	0.3929	0.0	0.3929
0.0253	21.875	700	0.0262	0.2048	0.4095	0.4095	nan	0.4095	0.0	0.4095
0.0274	22.5	720	0.0251	0.2201	0.4401	0.4401	nan	0.4401	0.0	0.4401
0.0259	23.125	740	0.0254	0.2410	0.4820	0.4820	nan	0.4820	0.0	0.4820
0.0211	23.75	760	0.0253	0.2554	0.5109	0.5109	nan	0.5109	0.0	0.5109
0.0172	24.375	780	0.0244	0.2268	0.4535	0.4535	nan	0.4535	0.0	0.4535
0.0229	25.0	800	0.0250	0.2352	0.4705	0.4705	nan	0.4705	0.0	0.4705
0.0189	25.625	820	0.0238	0.1980	0.3959	0.3959	nan	0.3959	0.0	0.3959
0.0263	26.25	840	0.0240	0.1846	0.3693	0.3693	nan	0.3693	0.0	0.3693
0.0155	26.875	860	0.0237	0.2297	0.4594	0.4594	nan	0.4594	0.0	0.4594
0.0229	27.5	880	0.0228	0.2472	0.4945	0.4945	nan	0.4945	0.0	0.4945
0.0205	28.125	900	0.0232	0.2453	0.4906	0.4906	nan	0.4906	0.0	0.4906
0.0193	28.75	920	0.0229	0.2524	0.5048	0.5048	nan	0.5048	0.0	0.5048
0.0222	29.375	940	0.0224	0.2806	0.5613	0.5613	nan	0.5613	0.0	0.5613
0.0231	30.0	960	0.0222	0.2717	0.5434	0.5434	nan	0.5434	0.0	0.5434
0.0224	30.625	980	0.0220	0.2446	0.4892	0.4892	nan	0.4892	0.0	0.4892
0.0269	31.25	1000	0.0220	0.2588	0.5176	0.5176	nan	0.5176	0.0	0.5176
0.0229	31.875	1020	0.0227	0.2735	0.5469	0.5469	nan	0.5469	0.0	0.5469
0.0211	32.5	1040	0.0218	0.2326	0.4652	0.4652	nan	0.4652	0.0	0.4652
0.0212	33.125	1060	0.0215	0.2690	0.5381	0.5381	nan	0.5381	0.0	0.5381
0.0197	33.75	1080	0.0213	0.2471	0.4943	0.4943	nan	0.4943	0.0	0.4943
0.0206	34.375	1100	0.0212	0.2534	0.5068	0.5068	nan	0.5068	0.0	0.5068
0.0216	35.0	1120	0.0214	0.2622	0.5245	0.5245	nan	0.5245	0.0	0.5245
0.0176	35.625	1140	0.0209	0.2574	0.5148	0.5148	nan	0.5148	0.0	0.5148
0.0158	36.25	1160	0.0209	0.2531	0.5062	0.5062	nan	0.5062	0.0	0.5062
0.0154	36.875	1180	0.0209	0.2457	0.4914	0.4914	nan	0.4914	0.0	0.4914
0.0117	37.5	1200	0.0207	0.2501	0.5003	0.5003	nan	0.5003	0.0	0.5003
0.0131	38.125	1220	0.0206	0.2701	0.5401	0.5401	nan	0.5401	0.0	0.5401
0.0216	38.75	1240	0.0207	0.2723	0.5446	0.5446	nan	0.5446	0.0	0.5446
0.0162	39.375	1260	0.0206	0.2534	0.5067	0.5067	nan	0.5067	0.0	0.5067
0.0237	40.0	1280	0.0206	0.2787	0.5573	0.5573	nan	0.5573	0.0	0.5573
0.0187	40.625	1300	0.0203	0.2649	0.5298	0.5298	nan	0.5298	0.0	0.5298
0.0136	41.25	1320	0.0203	0.2633	0.5266	0.5266	nan	0.5266	0.0	0.5266
0.015	41.875	1340	0.0204	0.2603	0.5207	0.5207	nan	0.5207	0.0	0.5207
0.0218	42.5	1360	0.0204	0.2685	0.5370	0.5370	nan	0.5370	0.0	0.5370
0.0236	43.125	1380	0.0201	0.2735	0.5471	0.5471	nan	0.5471	0.0	0.5471
0.0181	43.75	1400	0.0203	0.2609	0.5217	0.5217	nan	0.5217	0.0	0.5217
0.0103	44.375	1420	0.0201	0.2730	0.5460	0.5460	nan	0.5460	0.0	0.5460
0.0125	45.0	1440	0.0202	0.2651	0.5301	0.5301	nan	0.5301	0.0	0.5301

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0