segformer-b1-GFB

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: 0.4646
• Mean Iou: 0.6980
• Mean Accuracy: 0.7985
• Overall Accuracy: 0.9199
• Accuracy Unlabeled: 0.9630
• Accuracy Gbm: 0.8245
• Accuracy Podo: 0.7482
• Accuracy Endo: 0.6582
• Iou Unlabeled: 0.9177
• Iou Gbm: 0.7081
• Iou Podo: 0.6223
• Iou Endo: 0.5438

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• 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: 10
• 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
0.2783	1.0989	100	0.3191	0.5500	0.6348	0.8806	0.9690	0.6768	0.5333	0.3602	0.8818	0.5838	0.4362	0.2981
0.2116	2.1978	200	0.3145	0.5560	0.6580	0.8799	0.9503	0.8048	0.5330	0.3436	0.8833	0.5831	0.4525	0.3052
0.2333	3.2967	300	0.2593	0.6270	0.7281	0.9011	0.9627	0.7598	0.6617	0.5283	0.9021	0.6547	0.5276	0.4237
0.2253	4.3956	400	0.2582	0.6409	0.7774	0.9008	0.9374	0.8742	0.7192	0.5788	0.9022	0.6586	0.5714	0.4313
0.1657	5.4945	500	0.3006	0.6376	0.7280	0.9030	0.9710	0.6997	0.6728	0.5683	0.9004	0.6281	0.5650	0.4570
0.1814	6.5934	600	0.2519	0.6613	0.7745	0.9075	0.9518	0.8269	0.7246	0.5946	0.9056	0.6695	0.5892	0.4809
0.284	7.6923	700	0.2642	0.6567	0.7587	0.9082	0.9649	0.7735	0.6733	0.6230	0.9070	0.6609	0.5749	0.4842
0.1474	8.7912	800	0.2504	0.6612	0.7811	0.9085	0.9481	0.8704	0.7133	0.5927	0.9097	0.6758	0.5816	0.4776
0.1292	9.8901	900	0.2652	0.6621	0.7727	0.9084	0.9558	0.8163	0.7057	0.6132	0.9073	0.6753	0.5757	0.4899
0.1569	10.9890	1000	0.2582	0.6681	0.8105	0.9072	0.9406	0.8722	0.7224	0.7070	0.9063	0.6834	0.5882	0.4947
0.1352	12.0879	1100	0.2709	0.6750	0.7844	0.9133	0.9568	0.8251	0.7424	0.6131	0.9120	0.6829	0.6125	0.4927
0.0956	13.1868	1200	0.2897	0.6627	0.7500	0.9134	0.9670	0.8266	0.6794	0.5269	0.9127	0.6863	0.5835	0.4684
0.0922	14.2857	1300	0.2885	0.6842	0.7919	0.9160	0.9595	0.8328	0.7270	0.6482	0.9143	0.7005	0.6082	0.5139
0.1208	15.3846	1400	0.2828	0.6777	0.7899	0.9134	0.9587	0.8136	0.7226	0.6646	0.9117	0.6878	0.6032	0.5082
0.0951	16.4835	1500	0.3012	0.6705	0.7758	0.9114	0.9571	0.8140	0.7412	0.5908	0.9100	0.6753	0.6014	0.4953
0.1305	17.5824	1600	0.3082	0.6750	0.7871	0.9109	0.9564	0.7949	0.7425	0.6546	0.9080	0.6808	0.5958	0.5155
0.0844	18.6813	1700	0.2803	0.6841	0.7985	0.9144	0.9533	0.8252	0.7789	0.6365	0.9121	0.6971	0.6141	0.5130
0.0838	19.7802	1800	0.2975	0.6812	0.7853	0.9148	0.9585	0.8173	0.7563	0.6091	0.9125	0.6905	0.6150	0.5067
0.0883	20.8791	1900	0.3026	0.6821	0.7983	0.9135	0.9570	0.8001	0.7487	0.6874	0.9123	0.6876	0.5991	0.5295
0.0831	21.9780	2000	0.3038	0.6779	0.8143	0.9081	0.9384	0.8633	0.7737	0.6819	0.9044	0.6866	0.6012	0.5192
0.0771	23.0769	2100	0.2994	0.6818	0.7862	0.9152	0.9596	0.8046	0.7661	0.6146	0.9134	0.6917	0.6132	0.5090
0.0801	24.1758	2200	0.3132	0.6868	0.7923	0.9171	0.9579	0.8515	0.7399	0.6200	0.9156	0.7053	0.6123	0.5142
0.1058	25.2747	2300	0.3076	0.6868	0.8040	0.9143	0.9529	0.8166	0.7835	0.6631	0.9123	0.6933	0.6135	0.5280
0.0904	26.3736	2400	0.3152	0.6804	0.8201	0.9089	0.9380	0.8586	0.7875	0.6961	0.9056	0.6929	0.6000	0.5232
0.1197	27.4725	2500	0.3155	0.6866	0.7851	0.9171	0.9639	0.8077	0.7406	0.6283	0.9155	0.6966	0.6113	0.5230
0.0624	28.5714	2600	0.3164	0.6911	0.7993	0.9177	0.9628	0.8108	0.7286	0.6951	0.9159	0.6997	0.6148	0.5339
0.0782	29.6703	2700	0.3248	0.6941	0.7915	0.9191	0.9657	0.8091	0.7348	0.6564	0.9171	0.7056	0.6124	0.5415
0.0659	30.7692	2800	0.3203	0.6848	0.7859	0.9163	0.9635	0.8224	0.7055	0.6523	0.9141	0.6986	0.6017	0.5248
0.081	31.8681	2900	0.3249	0.6819	0.8141	0.9105	0.9402	0.8811	0.7664	0.6688	0.9081	0.6858	0.6093	0.5243
0.0851	32.9670	3000	0.3988	0.6659	0.7493	0.9119	0.9736	0.7579	0.6657	0.6002	0.9089	0.6686	0.5810	0.5053
0.0779	34.0659	3100	0.3361	0.6907	0.7972	0.9162	0.9584	0.8045	0.7768	0.6490	0.9135	0.6946	0.6192	0.5355
0.0626	35.1648	3200	0.3443	0.6860	0.7813	0.9173	0.9652	0.8134	0.7256	0.6210	0.9151	0.6949	0.6139	0.5203
0.0709	36.2637	3300	0.3250	0.6963	0.8080	0.9184	0.9560	0.8429	0.7673	0.6658	0.9166	0.7064	0.6243	0.5381
0.0661	37.3626	3400	0.3290	0.6919	0.7925	0.9187	0.9626	0.8309	0.7357	0.6407	0.9170	0.7068	0.6143	0.5295
0.0825	38.4615	3500	0.3347	0.6951	0.8019	0.9184	0.9568	0.8501	0.7592	0.6416	0.9163	0.7086	0.6203	0.5351
0.0668	39.5604	3600	0.3523	0.6860	0.8081	0.9133	0.9496	0.8334	0.7746	0.6750	0.9105	0.6947	0.6115	0.5272
0.0604	40.6593	3700	0.4000	0.6831	0.7693	0.9167	0.9704	0.7860	0.7129	0.6079	0.9137	0.6886	0.6102	0.5201
0.0913	41.7582	3800	0.3509	0.6967	0.8120	0.9179	0.9564	0.8277	0.7661	0.6977	0.9157	0.7064	0.6206	0.5442
0.045	42.8571	3900	0.3374	0.6989	0.8258	0.9170	0.9476	0.8647	0.7833	0.7077	0.9143	0.7103	0.6251	0.5460
0.0543	43.9560	4000	0.3796	0.6868	0.7833	0.9165	0.9672	0.7774	0.7326	0.6558	0.9137	0.6891	0.6110	0.5335
0.0621	45.0549	4100	0.3530	0.7017	0.8086	0.9206	0.9600	0.8345	0.7633	0.6768	0.9185	0.7138	0.6284	0.5460
0.0583	46.1538	4200	0.3672	0.6901	0.8045	0.9157	0.9561	0.8251	0.7518	0.6849	0.9136	0.6983	0.6111	0.5373
0.0464	47.2527	4300	0.3885	0.6952	0.7966	0.9197	0.9629	0.8376	0.7309	0.6550	0.9181	0.7113	0.6149	0.5364
0.056	48.3516	4400	0.3900	0.6950	0.7988	0.9186	0.9618	0.8124	0.7571	0.6638	0.9161	0.7044	0.6217	0.5380
0.0673	49.4505	4500	0.3864	0.6972	0.8041	0.9191	0.9590	0.8442	0.7472	0.6658	0.9172	0.7100	0.6195	0.5420
0.0402	50.5495	4600	0.3809	0.6814	0.7824	0.9148	0.9626	0.8002	0.7315	0.6352	0.9124	0.6897	0.6041	0.5192
0.0657	51.6484	4700	0.3789	0.6942	0.7968	0.9184	0.9614	0.8354	0.7313	0.6591	0.9164	0.7042	0.6145	0.5418
0.053	52.7473	4800	0.3832	0.6972	0.8079	0.9185	0.9570	0.8385	0.7625	0.6736	0.9163	0.7077	0.6225	0.5424
0.052	53.8462	4900	0.4184	0.6940	0.7909	0.9191	0.9689	0.7788	0.7386	0.6775	0.9171	0.6958	0.6198	0.5433
0.0572	54.9451	5000	0.3882	0.6993	0.7984	0.9204	0.9635	0.8270	0.7469	0.6563	0.9182	0.7085	0.6256	0.5451
0.0364	56.0440	5100	0.4099	0.6938	0.7945	0.9183	0.9643	0.8079	0.7345	0.6715	0.9157	0.6994	0.6171	0.5429
0.0443	57.1429	5200	0.3932	0.6974	0.7955	0.9200	0.9651	0.8207	0.7340	0.6622	0.9179	0.7090	0.6181	0.5446
0.0392	58.2418	5300	0.4024	0.6981	0.8006	0.9195	0.9617	0.8292	0.7479	0.6638	0.9173	0.7089	0.6205	0.5456
0.0378	59.3407	5400	0.4137	0.6993	0.7984	0.9205	0.9641	0.8240	0.7460	0.6595	0.9181	0.7122	0.6234	0.5433
0.0603	60.4396	5500	0.4196	0.6980	0.7950	0.9204	0.9656	0.8190	0.7379	0.6577	0.9181	0.7107	0.6200	0.5433
0.0511	61.5385	5600	0.4441	0.6891	0.7772	0.9179	0.9703	0.7892	0.7124	0.6368	0.9151	0.6931	0.6085	0.5397
0.059	62.6374	5700	0.4338	0.6955	0.7906	0.9197	0.9665	0.8110	0.7354	0.6493	0.9177	0.7035	0.6185	0.5423
0.0401	63.7363	5800	0.4535	0.6923	0.7847	0.9190	0.9689	0.7952	0.7266	0.6481	0.9168	0.7007	0.6126	0.5391
0.0334	64.8352	5900	0.4412	0.6958	0.7930	0.9196	0.9657	0.8180	0.7293	0.6589	0.9176	0.7044	0.6178	0.5432
0.0603	65.9341	6000	0.4229	0.6967	0.7923	0.9199	0.9658	0.8163	0.7361	0.6507	0.9176	0.7072	0.6183	0.5437
0.0453	67.0330	6100	0.4471	0.6972	0.7939	0.9201	0.9671	0.8057	0.7344	0.6682	0.9180	0.7054	0.6183	0.5473
0.0469	68.1319	6200	0.4214	0.6989	0.8029	0.9201	0.9622	0.8315	0.7421	0.6759	0.9182	0.7093	0.6230	0.5453
0.0416	69.2308	6300	0.4360	0.6972	0.7947	0.9203	0.9640	0.8321	0.7399	0.6428	0.9183	0.7107	0.6214	0.5383
0.0658	70.3297	6400	0.4419	0.6966	0.7984	0.9192	0.9626	0.8207	0.7464	0.6641	0.9169	0.7055	0.6205	0.5436
0.0572	71.4286	6500	0.4392	0.6981	0.7993	0.9197	0.9627	0.8271	0.7430	0.6644	0.9175	0.7084	0.6206	0.5457
0.0434	72.5275	6600	0.4521	0.6969	0.7982	0.9193	0.9639	0.8106	0.7464	0.6719	0.9171	0.7034	0.6209	0.5460
0.0396	73.6264	6700	0.4341	0.6983	0.7983	0.9199	0.9628	0.8280	0.7454	0.6569	0.9176	0.7072	0.6231	0.5453
0.0483	74.7253	6800	0.4324	0.6958	0.8038	0.9181	0.9571	0.8443	0.7553	0.6586	0.9157	0.7058	0.6211	0.5408
0.0411	75.8242	6900	0.4290	0.6980	0.8069	0.9186	0.9571	0.8391	0.7655	0.6660	0.9161	0.7079	0.6243	0.5435
0.0522	76.9231	7000	0.4583	0.6962	0.7916	0.9200	0.9652	0.8235	0.7378	0.6400	0.9177	0.7074	0.6222	0.5376
0.0452	78.0220	7100	0.4529	0.6979	0.7992	0.9198	0.9624	0.8266	0.7499	0.6578	0.9176	0.7081	0.6239	0.5420
0.0412	79.1209	7200	0.4497	0.6984	0.7964	0.9203	0.9650	0.8146	0.7471	0.6590	0.9181	0.7074	0.6235	0.5446
0.0421	80.2198	7300	0.4573	0.6975	0.7946	0.9203	0.9651	0.8209	0.7407	0.6518	0.9181	0.7086	0.6215	0.5418
0.0504	81.3187	7400	0.4639	0.6973	0.7952	0.9201	0.9641	0.8224	0.7478	0.6466	0.9181	0.7083	0.6232	0.5398
0.0318	82.4176	7500	0.4477	0.6996	0.8017	0.9202	0.9625	0.8255	0.7511	0.6675	0.9182	0.7080	0.6245	0.5478
0.0497	83.5165	7600	0.4687	0.6962	0.7914	0.9197	0.9664	0.8093	0.7372	0.6529	0.9175	0.7024	0.6205	0.5442
0.0779	84.6154	7700	0.4793	0.6955	0.7895	0.9199	0.9667	0.8153	0.7313	0.6445	0.9177	0.7051	0.6187	0.5405
0.035	85.7143	7800	0.4690	0.6982	0.7966	0.9202	0.9648	0.8173	0.7441	0.6600	0.9181	0.7064	0.6231	0.5451
0.0503	86.8132	7900	0.4654	0.6975	0.7981	0.9198	0.9628	0.8272	0.7459	0.6566	0.9177	0.7074	0.6222	0.5427
0.0491	87.9121	8000	0.4481	0.6986	0.8019	0.9195	0.9608	0.8266	0.7607	0.6596	0.9172	0.7076	0.6241	0.5454
0.0347	89.0110	8100	0.4530	0.6988	0.8005	0.9199	0.9620	0.8283	0.7528	0.6588	0.9177	0.7087	0.6238	0.5449
0.0445	90.1099	8200	0.4635	0.6976	0.7962	0.9201	0.9636	0.8271	0.7450	0.6490	0.9180	0.7091	0.6227	0.5407
0.0351	91.2088	8300	0.4566	0.6983	0.8010	0.9198	0.9617	0.8299	0.7498	0.6626	0.9176	0.7089	0.6225	0.5444
0.0379	92.3077	8400	0.4639	0.6977	0.7976	0.9199	0.9634	0.8245	0.7456	0.6570	0.9178	0.7079	0.6222	0.5431
0.0432	93.4066	8500	0.4643	0.6974	0.7961	0.9200	0.9637	0.8256	0.7440	0.6514	0.9178	0.7082	0.6218	0.5417
0.0396	94.5055	8600	0.4699	0.6972	0.7951	0.9200	0.9643	0.8214	0.7431	0.6517	0.9178	0.7074	0.6215	0.5420
0.0427	95.6044	8700	0.4709	0.6973	0.7957	0.9199	0.9643	0.8199	0.7439	0.6548	0.9177	0.7075	0.6214	0.5427
0.0469	96.7033	8800	0.4671	0.6976	0.7971	0.9199	0.9635	0.8251	0.7436	0.6564	0.9177	0.7081	0.6213	0.5433
0.0421	97.8022	8900	0.4606	0.6983	0.7991	0.9199	0.9628	0.8259	0.7483	0.6595	0.9178	0.7084	0.6224	0.5444
0.0579	98.9011	9000	0.4639	0.6981	0.7987	0.9199	0.9630	0.8249	0.7472	0.6598	0.9177	0.7082	0.6221	0.5443
0.0322	100.0	9100	0.4646	0.6980	0.7985	0.9199	0.9630	0.8245	0.7482	0.6582	0.9177	0.7081	0.6223	0.5438

Framework versions

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