segformer-b1-GFB-NF-512-1024

This model is a fine-tuned version of nvidia/mit-b1 on the segments/GFB dataset. It achieves the following results on the evaluation set:

• Loss: 0.6972
• Mean Iou: 0.5572
• Mean Accuracy: 0.6875
• Overall Accuracy: 0.8653
• Accuracy Unlabeled: 0.9296
• Accuracy Gbm: 0.7701
• Accuracy Podo: 0.6031
• Accuracy Endo: 0.4472
• Iou Unlabeled: 0.8707
• Iou Gbm: 0.5719
• Iou Podo: 0.4445
• Iou Endo: 0.3417

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 250
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Gbm	Accuracy Podo	Accuracy Endo	Iou Unlabeled	Iou Gbm	Iou Podo	Iou Endo
0.9456	1.0526	100	1.0895	0.3047	0.4705	0.7199	0.7850	0.9252	0.1442	0.0277	0.7493	0.3392	0.1054	0.0249
0.5926	2.1053	200	0.6151	0.3885	0.4834	0.8305	0.9514	0.7433	0.2228	0.0161	0.8441	0.5056	0.1886	0.0157
0.5148	3.1579	300	0.4627	0.4671	0.5747	0.8473	0.9395	0.7066	0.5266	0.1261	0.8597	0.5320	0.3614	0.1153
0.2959	4.2105	400	0.4571	0.4587	0.5350	0.8429	0.9715	0.5666	0.3330	0.2688	0.8439	0.4839	0.2852	0.2218
0.3072	5.2632	500	0.4246	0.5276	0.6815	0.8472	0.9031	0.8183	0.5609	0.4435	0.8563	0.5431	0.3966	0.3143
0.2677	6.3158	600	0.3923	0.5335	0.6337	0.8652	0.9540	0.6837	0.5495	0.3475	0.8705	0.5577	0.4148	0.2911
0.2468	7.3684	700	0.4119	0.5445	0.6949	0.8527	0.9082	0.7640	0.6493	0.4583	0.8589	0.5516	0.4257	0.3420
0.2456	8.4211	800	0.4212	0.5478	0.6944	0.8575	0.9136	0.7821	0.6369	0.4451	0.8649	0.5666	0.4297	0.3298
0.2206	9.4737	900	0.4098	0.5433	0.6663	0.8621	0.9381	0.7338	0.5410	0.4524	0.8688	0.5613	0.4090	0.3341
0.183	10.5263	1000	0.4265	0.5421	0.6582	0.8646	0.9414	0.7649	0.5134	0.4130	0.8719	0.5699	0.3997	0.3271
0.1808	11.5789	1100	0.4287	0.5491	0.6767	0.8635	0.9304	0.7711	0.5861	0.4192	0.8697	0.5709	0.4311	0.3246
0.1513	12.6316	1200	0.4276	0.5436	0.6613	0.8631	0.9381	0.7362	0.5696	0.4015	0.8686	0.5608	0.4240	0.3210
0.1532	13.6842	1300	0.4409	0.5557	0.6879	0.8632	0.9293	0.7460	0.6066	0.4698	0.8685	0.5640	0.4393	0.3510
0.1929	14.7368	1400	0.4531	0.5416	0.6609	0.8613	0.9425	0.6989	0.5376	0.4648	0.8667	0.5474	0.4129	0.3396
0.1722	15.7895	1500	0.4607	0.5497	0.6938	0.8581	0.9177	0.7781	0.6006	0.4786	0.8634	0.5615	0.4317	0.3424
0.0925	16.8421	1600	0.4733	0.5450	0.6895	0.8571	0.9143	0.8035	0.5993	0.4409	0.8643	0.5607	0.4282	0.3267
0.1685	17.8947	1700	0.4764	0.5487	0.6792	0.8600	0.9319	0.7100	0.5906	0.4843	0.8639	0.5579	0.4245	0.3483
0.1369	18.9474	1800	0.4843	0.5507	0.6834	0.8620	0.9303	0.7545	0.5686	0.4803	0.8679	0.5658	0.4236	0.3455
0.1281	20.0	1900	0.4728	0.5561	0.7008	0.8607	0.9189	0.7776	0.6160	0.4905	0.8657	0.5721	0.4380	0.3487
0.1207	21.0526	2000	0.4867	0.5523	0.6780	0.8645	0.9329	0.7500	0.6021	0.4269	0.8699	0.5676	0.4401	0.3316
0.1395	22.1053	2100	0.4914	0.5530	0.6786	0.8655	0.9322	0.7766	0.5858	0.4198	0.8721	0.5723	0.4354	0.3322
0.137	23.1579	2200	0.5332	0.5431	0.6948	0.8533	0.9042	0.8170	0.6288	0.4291	0.8589	0.5581	0.4324	0.3229
0.1078	24.2105	2300	0.5052	0.5534	0.6923	0.8618	0.9264	0.7551	0.5925	0.4952	0.8667	0.5673	0.4393	0.3404
0.1127	25.2632	2400	0.5219	0.5508	0.6767	0.8632	0.9323	0.7490	0.5894	0.4363	0.8670	0.5696	0.4349	0.3319
0.1172	26.3158	2500	0.5395	0.5473	0.6888	0.8580	0.9230	0.7481	0.5894	0.4945	0.8633	0.5542	0.4329	0.3387
0.0976	27.3684	2600	0.5311	0.5497	0.6793	0.8625	0.9306	0.7600	0.5756	0.4510	0.8683	0.5668	0.4268	0.3371
0.1128	28.4211	2700	0.5346	0.5423	0.6641	0.8613	0.9365	0.7460	0.5340	0.4401	0.8665	0.5617	0.4082	0.3329
0.0711	29.4737	2800	0.5315	0.5531	0.6854	0.8625	0.9283	0.7466	0.6137	0.4528	0.8678	0.5624	0.4421	0.3401
0.0795	30.5263	2900	0.5507	0.5564	0.7016	0.8608	0.9162	0.7929	0.6300	0.4672	0.8655	0.5718	0.4485	0.3397
0.1243	31.5789	3000	0.5455	0.5575	0.6928	0.8646	0.9284	0.7676	0.5984	0.4769	0.8707	0.5770	0.4363	0.3460
0.1078	32.6316	3100	0.5397	0.5566	0.6921	0.8632	0.9289	0.7477	0.5979	0.4938	0.8681	0.5693	0.4373	0.3518
0.0867	33.6842	3200	0.5545	0.5584	0.6949	0.8643	0.9251	0.7761	0.6175	0.4611	0.8698	0.5731	0.4478	0.3431
0.1081	34.7368	3300	0.5504	0.5561	0.6899	0.8637	0.9266	0.7720	0.6050	0.4559	0.8693	0.5724	0.4379	0.3446
0.0761	35.7895	3400	0.5615	0.5504	0.6774	0.8633	0.9310	0.7592	0.5920	0.4273	0.8687	0.5666	0.4357	0.3307
0.0858	36.8421	3500	0.5828	0.5496	0.6765	0.8629	0.9311	0.7647	0.5763	0.4338	0.8685	0.5638	0.4307	0.3353
0.0761	37.8947	3600	0.5640	0.5507	0.6736	0.8638	0.9353	0.7406	0.5807	0.4379	0.8677	0.5681	0.4318	0.3351
0.0808	38.9474	3700	0.5691	0.5508	0.6771	0.8643	0.9312	0.7680	0.5937	0.4154	0.8695	0.5713	0.4391	0.3235
0.149	40.0	3800	0.5780	0.5484	0.6748	0.8623	0.9333	0.7538	0.5556	0.4565	0.8678	0.5639	0.4185	0.3434
0.0779	41.0526	3900	0.5739	0.5545	0.6942	0.8619	0.9210	0.7903	0.6081	0.4576	0.8672	0.5704	0.4442	0.3361
0.0839	42.1053	4000	0.5863	0.5541	0.6884	0.8632	0.9255	0.7768	0.6065	0.4447	0.8686	0.5696	0.4436	0.3346
0.1068	43.1579	4100	0.5705	0.5573	0.7025	0.8615	0.9186	0.7766	0.6337	0.4812	0.8672	0.5693	0.4477	0.3450
0.0843	44.2105	4200	0.5941	0.5583	0.6935	0.8647	0.9262	0.7757	0.6139	0.4580	0.8697	0.5747	0.4494	0.3394
0.0657	45.2632	4300	0.5937	0.5565	0.6929	0.8627	0.9239	0.7592	0.6378	0.4509	0.8675	0.5679	0.4506	0.3400
0.0651	46.3158	4400	0.6141	0.5527	0.6814	0.8638	0.9297	0.7756	0.5812	0.4390	0.8690	0.5693	0.4356	0.3369
0.0922	47.3684	4500	0.6089	0.5558	0.6848	0.8645	0.9311	0.7540	0.6024	0.4519	0.8693	0.5713	0.4401	0.3425
0.0528	48.4211	4600	0.6147	0.5529	0.6852	0.8628	0.9249	0.7833	0.6074	0.4252	0.8677	0.5679	0.4453	0.3307
0.0739	49.4737	4700	0.6108	0.5592	0.6998	0.8626	0.9216	0.7692	0.6309	0.4774	0.8674	0.5705	0.4471	0.3519
0.1017	50.5263	4800	0.6339	0.5526	0.6784	0.8644	0.9327	0.7681	0.5707	0.4421	0.8691	0.5689	0.4341	0.3382
0.0791	51.5789	4900	0.6122	0.5523	0.6827	0.8627	0.9291	0.7638	0.5851	0.4530	0.8673	0.5657	0.4370	0.3390
0.0643	52.6316	5000	0.6261	0.5530	0.6855	0.8630	0.9265	0.7765	0.5970	0.4420	0.8681	0.5669	0.4420	0.3351
0.0701	53.6842	5100	0.6176	0.5572	0.6937	0.8634	0.9248	0.7692	0.6202	0.4605	0.8685	0.5712	0.4467	0.3422
0.0665	54.7368	5200	0.6148	0.5555	0.6853	0.8649	0.9295	0.7793	0.5881	0.4443	0.8704	0.5720	0.4397	0.3399
0.0954	55.7895	5300	0.6239	0.5540	0.6801	0.8653	0.9331	0.7636	0.5856	0.4382	0.8704	0.5719	0.4396	0.3339
0.0642	56.8421	5400	0.6293	0.5567	0.6893	0.8639	0.9271	0.7679	0.6107	0.4517	0.8688	0.5681	0.4462	0.3435
0.069	57.8947	5500	0.6354	0.5569	0.6906	0.8638	0.9264	0.7713	0.6087	0.4560	0.8691	0.5694	0.4434	0.3456
0.068	58.9474	5600	0.6421	0.5561	0.6872	0.8646	0.9289	0.7702	0.6022	0.4477	0.8699	0.5705	0.4455	0.3384
0.1144	60.0	5700	0.6376	0.5549	0.6849	0.8645	0.9288	0.7700	0.6082	0.4326	0.8698	0.5710	0.4456	0.3333
0.0746	61.0526	5800	0.6433	0.5531	0.6755	0.8658	0.9351	0.7633	0.5809	0.4229	0.8709	0.5707	0.4390	0.3317
0.0702	62.1053	5900	0.6516	0.5549	0.6819	0.8654	0.9330	0.7613	0.5878	0.4454	0.8709	0.5705	0.4393	0.3388
0.0739	63.1579	6000	0.6511	0.5549	0.6830	0.8647	0.9322	0.7570	0.5895	0.4534	0.8699	0.5672	0.4404	0.3419
0.0733	64.2105	6100	0.6616	0.5571	0.6894	0.8651	0.9288	0.7727	0.6018	0.4542	0.8708	0.5712	0.4454	0.3411
0.0943	65.2632	6200	0.6595	0.5572	0.6877	0.8653	0.9292	0.7760	0.5998	0.4459	0.8709	0.5732	0.4434	0.3415
0.0826	66.3158	6300	0.6623	0.5553	0.6850	0.8646	0.9299	0.7634	0.6043	0.4426	0.8699	0.5700	0.4432	0.3380
0.0584	67.3684	6400	0.6611	0.5535	0.6858	0.8634	0.9277	0.7715	0.5960	0.4481	0.8688	0.5680	0.4415	0.3355
0.0613	68.4211	6500	0.6670	0.5573	0.6960	0.8631	0.9220	0.7861	0.6184	0.4574	0.8685	0.5709	0.4474	0.3423
0.0885	69.4737	6600	0.6687	0.5564	0.6853	0.8652	0.9308	0.7639	0.6018	0.4449	0.8703	0.5720	0.4433	0.3400
0.0696	70.5263	6700	0.6711	0.5574	0.6920	0.8642	0.9268	0.7741	0.6035	0.4637	0.8699	0.5710	0.4432	0.3456
0.057	71.5789	6800	0.6672	0.5566	0.6884	0.8646	0.9284	0.7711	0.6034	0.4507	0.8698	0.5717	0.4433	0.3417
0.0516	72.6316	6900	0.6752	0.5557	0.6851	0.8651	0.9300	0.7669	0.6061	0.4374	0.8704	0.5708	0.4463	0.3354
0.0537	73.6842	7000	0.6708	0.5564	0.6846	0.8656	0.9319	0.7630	0.5967	0.4469	0.8709	0.5722	0.4425	0.3402
0.0741	74.7368	7100	0.6764	0.5575	0.6885	0.8652	0.9292	0.7663	0.6128	0.4456	0.8706	0.5728	0.4469	0.3398
0.0842	75.7895	7200	0.6727	0.5565	0.6897	0.8643	0.9266	0.7811	0.6025	0.4487	0.8697	0.5727	0.4431	0.3403
0.0849	76.8421	7300	0.6830	0.5574	0.6874	0.8656	0.9298	0.7738	0.6000	0.4459	0.8711	0.5726	0.4447	0.3414
0.0612	77.8947	7400	0.6802	0.5579	0.6856	0.8664	0.9327	0.7617	0.6011	0.4471	0.8718	0.5740	0.4446	0.3411
0.0738	78.9474	7500	0.6790	0.5590	0.6894	0.8660	0.9296	0.7704	0.6105	0.4472	0.8712	0.5736	0.4490	0.3419
0.0575	80.0	7600	0.6868	0.5583	0.6896	0.8655	0.9287	0.7763	0.6038	0.4494	0.8709	0.5730	0.4460	0.3432
0.0665	81.0526	7700	0.6792	0.5579	0.6904	0.8650	0.9278	0.7752	0.6071	0.4515	0.8704	0.5724	0.4463	0.3424
0.0725	82.1053	7800	0.6867	0.5580	0.6904	0.8650	0.9278	0.7690	0.6165	0.4484	0.8704	0.5718	0.4480	0.3418
0.074	83.1579	7900	0.6843	0.5573	0.6914	0.8642	0.9263	0.7739	0.6131	0.4524	0.8698	0.5710	0.4459	0.3424
0.0626	84.2105	8000	0.6923	0.5575	0.6894	0.8650	0.9284	0.7709	0.6069	0.4515	0.8705	0.5718	0.4447	0.3430
0.0667	85.2632	8100	0.6884	0.5572	0.6871	0.8653	0.9300	0.7723	0.5953	0.4507	0.8707	0.5720	0.4420	0.3441
0.048	86.3158	8200	0.6932	0.5572	0.6853	0.8658	0.9317	0.7651	0.5976	0.4466	0.8711	0.5722	0.4429	0.3426
0.0649	87.3684	8300	0.6918	0.5564	0.6839	0.8656	0.9318	0.7650	0.5979	0.4409	0.8708	0.5718	0.4432	0.3399
0.0602	88.4211	8400	0.6869	0.5582	0.6903	0.8651	0.9282	0.7721	0.6088	0.4522	0.8705	0.5723	0.4463	0.3439
0.0678	89.4737	8500	0.6896	0.5570	0.6863	0.8654	0.9305	0.7677	0.6021	0.4451	0.8708	0.5723	0.4442	0.3407
0.0661	90.5263	8600	0.6943	0.5575	0.6881	0.8653	0.9298	0.7687	0.6019	0.4522	0.8707	0.5725	0.4439	0.3430
0.0712	91.5789	8700	0.6962	0.5572	0.6882	0.8650	0.9289	0.7705	0.6067	0.4467	0.8704	0.5719	0.4450	0.3415
0.061	92.6316	8800	0.6919	0.5567	0.6873	0.8649	0.9291	0.7699	0.6050	0.4453	0.8703	0.5716	0.4443	0.3407
0.0665	93.6842	8900	0.6951	0.5562	0.6843	0.8655	0.9310	0.7697	0.5963	0.4403	0.8708	0.5719	0.4426	0.3397
0.0725	94.7368	9000	0.6954	0.5568	0.6866	0.8652	0.9298	0.7717	0.5997	0.4452	0.8707	0.5719	0.4435	0.3412
0.0612	95.7895	9100	0.6919	0.5571	0.6864	0.8654	0.9306	0.7672	0.6003	0.4477	0.8708	0.5720	0.4438	0.3418
0.0679	96.8421	9200	0.6985	0.5567	0.6854	0.8655	0.9309	0.7685	0.5982	0.4442	0.8709	0.5720	0.4432	0.3407
0.0712	97.8947	9300	0.6977	0.5568	0.6869	0.8652	0.9298	0.7705	0.6005	0.4470	0.8706	0.5716	0.4436	0.3414
0.0669	98.9474	9400	0.6917	0.5572	0.6871	0.8653	0.9300	0.7698	0.6007	0.4479	0.8708	0.5718	0.4439	0.3422
0.0714	100.0	9500	0.6972	0.5572	0.6875	0.8653	0.9296	0.7701	0.6031	0.4472	0.8707	0.5719	0.4445	0.3417

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.1+cu130
• Datasets 4.4.1
• Tokenizers 0.22.1