segformer-b1-GFB-NF

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: 0.6178
• Mean Iou: 0.5471
• Mean Accuracy: 0.6710
• Overall Accuracy: 0.8626
• Accuracy Unlabeled: 0.9347
• Accuracy Gbm: 0.7396
• Accuracy Podo: 0.5748
• Accuracy Endo: 0.4349
• Iou Unlabeled: 0.8679
• Iou Gbm: 0.5605
• Iou Podo: 0.4274
• Iou Endo: 0.3327

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
0.8664	2.1739	100	0.9327	0.3725	0.5216	0.7798	0.8537	0.8388	0.3476	0.0463	0.8014	0.4269	0.2206	0.0409
0.5846	4.3478	200	0.5418	0.4327	0.5234	0.8396	0.9558	0.6954	0.3076	0.1347	0.8518	0.5067	0.2559	0.1164
0.4017	6.5217	300	0.4348	0.5014	0.6032	0.8552	0.9528	0.6924	0.4408	0.3267	0.8638	0.5462	0.3467	0.2488
0.2923	8.6957	400	0.4294	0.4935	0.5802	0.8516	0.9657	0.5452	0.4955	0.3145	0.8566	0.4829	0.3677	0.2669
0.2786	10.8696	500	0.4279	0.5243	0.6746	0.8472	0.9044	0.7919	0.6048	0.3974	0.8539	0.5456	0.4163	0.2813
0.2292	13.0435	600	0.4181	0.5354	0.6563	0.8598	0.9366	0.7009	0.5995	0.3882	0.8668	0.5471	0.4258	0.3019
0.1743	15.2174	700	0.4424	0.5407	0.6923	0.8545	0.9144	0.7797	0.5785	0.4966	0.8615	0.5586	0.4223	0.3205
0.185	17.3913	800	0.4472	0.5383	0.6584	0.8611	0.9359	0.7542	0.5391	0.4043	0.8664	0.5601	0.4141	0.3126
0.1719	19.5652	900	0.4831	0.5291	0.6621	0.8539	0.9227	0.7689	0.5509	0.4059	0.8610	0.5517	0.4009	0.3027
0.1358	21.7391	1000	0.4655	0.5423	0.6763	0.8577	0.9245	0.7443	0.6020	0.4342	0.8625	0.5571	0.4289	0.3208
0.1272	23.9130	1100	0.4818	0.5435	0.6860	0.8566	0.9160	0.7723	0.6272	0.4283	0.8625	0.5558	0.4391	0.3167
0.1327	26.0870	1200	0.4968	0.5388	0.6590	0.8610	0.9373	0.7182	0.5750	0.4056	0.8660	0.5543	0.4285	0.3063
0.1408	28.2609	1300	0.5067	0.5426	0.6770	0.8580	0.9238	0.7691	0.5746	0.4405	0.8638	0.5529	0.4258	0.3280
0.1001	30.4348	1400	0.5186	0.5403	0.6650	0.8594	0.9308	0.7576	0.5549	0.4166	0.8642	0.5564	0.4179	0.3226
0.1259	32.6087	1500	0.5258	0.5492	0.6842	0.8603	0.9257	0.7632	0.5821	0.4660	0.8654	0.5638	0.4269	0.3405
0.0915	34.7826	1600	0.5144	0.5450	0.6718	0.8605	0.9331	0.7257	0.5784	0.4500	0.8655	0.5537	0.4266	0.3342
0.1066	36.9565	1700	0.5287	0.5485	0.6835	0.8601	0.9264	0.7572	0.5798	0.4706	0.8653	0.5624	0.4258	0.3404
0.0883	39.1304	1800	0.5622	0.5336	0.6492	0.8596	0.9384	0.7303	0.5413	0.3868	0.8642	0.5516	0.4120	0.3068
0.1505	41.3043	1900	0.5516	0.5471	0.6861	0.8585	0.9238	0.7428	0.6037	0.4740	0.8638	0.5559	0.4348	0.3338
0.0952	43.4783	2000	0.5529	0.5454	0.6733	0.8611	0.9299	0.7427	0.6019	0.4185	0.8666	0.5588	0.4346	0.3216
0.1029	45.6522	2100	0.5646	0.5437	0.6634	0.8613	0.9386	0.7127	0.5613	0.4409	0.8659	0.5516	0.4204	0.3369
0.0925	47.8261	2200	0.5642	0.5433	0.6743	0.8598	0.9265	0.7815	0.5573	0.4317	0.8657	0.5616	0.4186	0.3274
0.1156	50.0	2300	0.5629	0.5484	0.6778	0.8617	0.9319	0.7298	0.5946	0.4550	0.8674	0.5584	0.4326	0.3351
0.072	52.1739	2400	0.5716	0.5427	0.6591	0.8629	0.9404	0.7166	0.5721	0.4073	0.8681	0.5565	0.4254	0.3209
0.0873	54.3478	2500	0.5788	0.5448	0.6744	0.8601	0.9301	0.7406	0.5807	0.4462	0.8654	0.5587	0.4251	0.3301
0.0775	56.5217	2600	0.5829	0.5486	0.6755	0.8624	0.9321	0.7391	0.5976	0.4334	0.8678	0.5599	0.4369	0.3297
0.0808	58.6957	2700	0.5802	0.5476	0.6748	0.8623	0.9309	0.7552	0.5871	0.4261	0.8680	0.5638	0.4311	0.3275
0.0545	60.8696	2800	0.5893	0.5489	0.6766	0.8621	0.9315	0.7379	0.5983	0.4389	0.8672	0.5601	0.4353	0.3328
0.094	63.0435	2900	0.5932	0.5448	0.6645	0.8629	0.9369	0.7406	0.5665	0.4140	0.8678	0.5623	0.4254	0.3236
0.1134	65.2174	3000	0.6156	0.5394	0.6563	0.8614	0.9393	0.7223	0.5559	0.4078	0.8666	0.5530	0.4186	0.3195
0.0633	67.3913	3100	0.6059	0.5434	0.6637	0.8626	0.9356	0.7465	0.5708	0.4017	0.8676	0.5635	0.4257	0.3168
0.0891	69.5652	3200	0.6073	0.5459	0.6665	0.8633	0.9375	0.7382	0.5644	0.4261	0.8687	0.5620	0.4245	0.3284
0.0801	71.7391	3300	0.6087	0.5430	0.6607	0.8627	0.9391	0.7283	0.5620	0.4134	0.8677	0.5594	0.4220	0.3229
0.0714	73.9130	3400	0.6019	0.5459	0.6689	0.8626	0.9357	0.7435	0.5607	0.4356	0.8681	0.5615	0.4227	0.3314
0.0632	76.0870	3500	0.6144	0.5457	0.6664	0.8632	0.9367	0.7409	0.5689	0.4190	0.8685	0.5620	0.4264	0.3260
0.0733	78.2609	3600	0.6120	0.5458	0.6680	0.8626	0.9359	0.7385	0.5679	0.4298	0.8679	0.5604	0.4246	0.3302
0.0612	80.4348	3700	0.6143	0.5468	0.6718	0.8622	0.9330	0.7450	0.5798	0.4294	0.8676	0.5606	0.4286	0.3304
0.0546	82.6087	3800	0.6186	0.5453	0.6663	0.8627	0.9366	0.7356	0.5708	0.4221	0.8679	0.5603	0.4258	0.3274
0.0531	84.7826	3900	0.6160	0.5472	0.6709	0.8626	0.9343	0.7407	0.5789	0.4296	0.8678	0.5612	0.4285	0.3312
0.0646	86.9565	4000	0.6185	0.5465	0.6701	0.8625	0.9347	0.7425	0.5717	0.4315	0.8678	0.5607	0.4268	0.3308
0.0793	89.1304	4100	0.6166	0.5461	0.6690	0.8625	0.9347	0.7405	0.5767	0.4240	0.8677	0.5607	0.4278	0.3283
0.0661	91.3043	4200	0.6194	0.5465	0.6698	0.8625	0.9349	0.7406	0.5714	0.4325	0.8677	0.5607	0.4261	0.3313
0.0796	93.4783	4300	0.6170	0.5471	0.6715	0.8625	0.9339	0.7428	0.5773	0.4321	0.8679	0.5609	0.4281	0.3317
0.0625	95.6522	4400	0.6185	0.5468	0.6704	0.8625	0.9346	0.7407	0.5760	0.4303	0.8679	0.5606	0.4278	0.3310
0.0683	97.8261	4500	0.6215	0.5466	0.6699	0.8626	0.9348	0.7419	0.5739	0.4290	0.8680	0.5609	0.4270	0.3307
0.0611	100.0	4600	0.6178	0.5471	0.6710	0.8626	0.9347	0.7396	0.5748	0.4349	0.8679	0.5605	0.4274	0.3327

Framework versions

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