sagittal-b4-finetuned-segments

This model is a fine-tuned version of nvidia/mit-b4 on the jenniferlumeng/Sagittal dataset. It achieves the following results on the evaluation set:

• Loss: 0.5610
• Mean Iou: 0.6387
• Mean Accuracy: 0.7597
• Overall Accuracy: 0.7684
• Accuracy Background: nan
• Accuracy Olfactory bulb: 0.7170
• Accuracy Anterior olfactory nucleus: 0.6456
• Accuracy Basal ganglia: 0.7788
• Accuracy Cortex: 0.7965
• Accuracy Hypothalamus: 0.6187
• Accuracy Thalamus: 0.7553
• Accuracy Hippocampus: 0.8524
• Accuracy Midbrain: 0.8602
• Accuracy Cerebellum: 0.7899
• Accuracy Pons and medulla: 0.7831
• Iou Background: 0.0
• Iou Olfactory bulb: 0.6979
• Iou Anterior olfactory nucleus: 0.5897
• Iou Basal ganglia: 0.7036
• Iou Cortex: 0.7569
• Iou Hypothalamus: 0.5348
• Iou Thalamus: 0.7058
• Iou Hippocampus: 0.8192
• Iou Midbrain: 0.7187
• Iou Cerebellum: 0.7689
• Iou Pons and medulla: 0.7295

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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Olfactory bulb	Accuracy Anterior olfactory nucleus	Accuracy Basal ganglia	Accuracy Cortex	Accuracy Hypothalamus	Accuracy Thalamus	Accuracy Hippocampus	Accuracy Midbrain	Accuracy Cerebellum	Accuracy Pons and medulla	Iou Background	Iou Olfactory bulb	Iou Anterior olfactory nucleus	Iou Basal ganglia	Iou Cortex	Iou Hypothalamus	Iou Thalamus	Iou Hippocampus	Iou Midbrain	Iou Cerebellum	Iou Pons and medulla
1.1622	3.3333	20	1.5275	0.2290	0.2936	0.3032	nan	0.3546	0.0661	0.3629	0.4619	0.3534	0.0245	0.4997	0.3121	0.3472	0.1532	0.0	0.3179	0.0654	0.2793	0.3896	0.3249	0.0232	0.4386	0.1852	0.3458	0.1488
0.8211	6.6667	40	1.0160	0.3284	0.4622	0.4886	nan	0.4025	0.2457	0.4573	0.7211	0.3213	0.7259	0.5201	0.4266	0.3944	0.4069	0.0	0.3734	0.2422	0.3826	0.4803	0.2946	0.2820	0.4623	0.3025	0.3941	0.3986
0.2823	10.0	60	0.9503	0.4263	0.5468	0.5681	nan	0.4108	0.3905	0.5659	0.6721	0.4120	0.7734	0.5256	0.6114	0.6022	0.5039	0.0	0.4030	0.3792	0.4867	0.6002	0.3712	0.5518	0.4860	0.4593	0.4789	0.4733
0.4346	13.3333	80	0.6683	0.5384	0.6798	0.7221	nan	0.5424	0.5376	0.7562	0.8557	0.5587	0.8000	0.5245	0.7792	0.7022	0.7419	0.0	0.5198	0.5055	0.6668	0.7488	0.4901	0.6675	0.4835	0.5702	0.6572	0.6131
0.1348	16.6667	100	0.5909	0.5275	0.6836	0.7131	nan	0.5024	0.5379	0.6884	0.7820	0.6158	0.8733	0.5253	0.7972	0.8618	0.6525	0.0	0.4253	0.5029	0.5920	0.7553	0.5203	0.5693	0.4756	0.5773	0.7332	0.6511
0.1317	20.0	120	0.5279	0.6000	0.7499	0.7699	nan	0.6679	0.6691	0.6804	0.9212	0.6790	0.7882	0.7477	0.8093	0.8195	0.7169	0.0	0.6282	0.6283	0.6090	0.7796	0.6010	0.5967	0.6350	0.6556	0.7736	0.6933
0.2667	23.3333	140	0.6451	0.5482	0.6840	0.6961	nan	0.6738	0.5915	0.6175	0.7717	0.6215	0.7162	0.7077	0.7127	0.7174	0.7097	0.0	0.6220	0.5489	0.5643	0.7119	0.5204	0.5866	0.6468	0.5720	0.6582	0.5986
0.3673	26.6667	160	0.5395	0.5843	0.7265	0.7280	nan	0.7682	0.6859	0.6984	0.8040	0.6214	0.7752	0.8302	0.7929	0.5215	0.7669	0.0	0.7397	0.6224	0.6607	0.6138	0.5355	0.6739	0.6855	0.6733	0.5017	0.7208
0.345	30.0	180	0.4865	0.6101	0.7534	0.7675	nan	0.7244	0.7111	0.7634	0.9073	0.7027	0.7449	0.7589	0.8557	0.6596	0.7061	0.0	0.7089	0.6334	0.6898	0.6957	0.5837	0.6786	0.6832	0.7091	0.6404	0.6886
0.1892	33.3333	200	0.5088	0.6134	0.7589	0.7739	nan	0.6971	0.6785	0.7077	0.8255	0.6950	0.7285	0.8019	0.7823	0.8302	0.8419	0.0	0.6760	0.6139	0.6244	0.7471	0.5948	0.6243	0.7012	0.6364	0.7359	0.7934
0.283	36.6667	220	0.5012	0.6032	0.7387	0.7525	nan	0.6736	0.6548	0.6843	0.8329	0.6138	0.7489	0.8097	0.7708	0.8219	0.7763	0.0	0.6511	0.5898	0.5952	0.7460	0.5490	0.6433	0.7184	0.6573	0.7478	0.7373
0.3255	40.0	240	0.4538	0.6439	0.7751	0.7926	nan	0.6323	0.6450	0.7895	0.8253	0.6834	0.8150	0.8167	0.8587	0.8580	0.8274	0.0	0.6155	0.5910	0.6879	0.7771	0.5999	0.7198	0.7293	0.7568	0.8085	0.7969
0.148	43.3333	260	0.5867	0.5934	0.7219	0.7242	nan	0.5819	0.6130	0.7968	0.7211	0.6326	0.7201	0.8921	0.7792	0.7586	0.7235	0.0	0.5698	0.5527	0.7039	0.6853	0.5568	0.6735	0.7675	0.6665	0.6775	0.6738
0.2442	46.6667	280	0.5438	0.6123	0.7363	0.7502	nan	0.6327	0.6296	0.7893	0.7839	0.5792	0.7350	0.8244	0.8132	0.7980	0.7780	0.0	0.6221	0.5730	0.6917	0.7543	0.5004	0.6962	0.7619	0.6561	0.7618	0.7179
0.1645	50.0	300	0.5079	0.6323	0.7651	0.7711	nan	0.7346	0.6775	0.7749	0.7836	0.6132	0.7336	0.8661	0.8496	0.8220	0.7960	0.0	0.6891	0.6033	0.7091	0.7671	0.5359	0.6642	0.7340	0.7250	0.7986	0.7295
0.2699	53.3333	320	0.5663	0.6069	0.7401	0.7475	nan	0.7376	0.6604	0.7358	0.8071	0.6238	0.7225	0.8290	0.8199	0.7012	0.7635	0.0	0.7229	0.6041	0.6395	0.7020	0.5321	0.6502	0.7546	0.6855	0.6720	0.7131
0.2053	56.6667	340	0.5013	0.6341	0.7684	0.7750	nan	0.7147	0.6551	0.7489	0.8326	0.6458	0.8202	0.8792	0.8516	0.7662	0.7696	0.0	0.6918	0.5916	0.6512	0.7922	0.5612	0.7269	0.7414	0.7425	0.7519	0.7245
0.2427	60.0	360	0.4900	0.6275	0.7673	0.7721	nan	0.7584	0.7267	0.7405	0.8320	0.6785	0.7632	0.8677	0.8152	0.6697	0.8215	0.0	0.7289	0.6565	0.6647	0.7254	0.5795	0.6798	0.7799	0.6798	0.6329	0.7752
0.0668	63.3333	380	0.4845	0.6435	0.7722	0.7766	nan	0.7479	0.7064	0.7754	0.7830	0.6316	0.7340	0.8832	0.8429	0.7855	0.8320	0.0	0.7189	0.6336	0.6988	0.7412	0.5582	0.6887	0.8092	0.7069	0.7433	0.7797
0.1278	66.6667	400	0.5318	0.6220	0.7447	0.7560	nan	0.7063	0.6682	0.7959	0.7900	0.6057	0.7272	0.8067	0.8161	0.7418	0.7891	0.0	0.6939	0.6056	0.7106	0.7178	0.5353	0.7047	0.7581	0.6757	0.7074	0.7330
0.1184	70.0	420	0.5153	0.6434	0.7695	0.7778	nan	0.7200	0.6898	0.7627	0.8246	0.6589	0.7738	0.8395	0.8716	0.7698	0.7847	0.0	0.6858	0.6246	0.7008	0.7713	0.5730	0.6913	0.8064	0.7338	0.7483	0.7425
0.1317	73.3333	440	0.5403	0.6346	0.7586	0.7668	nan	0.7143	0.6677	0.7672	0.7990	0.5974	0.7354	0.8611	0.8529	0.8030	0.7876	0.0	0.6901	0.6051	0.6957	0.7751	0.5214	0.6903	0.8054	0.7020	0.7681	0.7279
0.0959	76.6667	460	0.5506	0.6325	0.7529	0.7596	nan	0.7081	0.6401	0.7706	0.7878	0.6339	0.7571	0.8553	0.8437	0.7636	0.7686	0.0	0.6859	0.5831	0.6937	0.7470	0.5459	0.7144	0.8041	0.7182	0.7359	0.7289
0.1181	80.0	480	0.5810	0.6227	0.7489	0.7528	nan	0.7194	0.6986	0.7478	0.7786	0.6016	0.7453	0.8501	0.8435	0.7140	0.7897	0.0	0.7035	0.6368	0.6793	0.7059	0.5201	0.6781	0.7990	0.7002	0.6957	0.7306
0.1272	83.3333	500	0.5927	0.6213	0.7406	0.7501	nan	0.7056	0.6515	0.7716	0.7891	0.6042	0.7289	0.8221	0.8266	0.7345	0.7725	0.0	0.6898	0.5965	0.7000	0.7362	0.5184	0.7017	0.7840	0.6860	0.7035	0.7186
0.1653	86.6667	520	0.5653	0.6368	0.7586	0.7645	nan	0.7195	0.6718	0.7697	0.7843	0.6201	0.7360	0.8479	0.8640	0.8044	0.7683	0.0	0.7011	0.6056	0.7009	0.7658	0.5286	0.7009	0.8013	0.7081	0.7787	0.7136
0.1633	90.0	540	0.5539	0.6421	0.7641	0.7693	nan	0.7257	0.6989	0.7533	0.8062	0.6249	0.7595	0.8532	0.8622	0.7844	0.7728	0.0	0.7107	0.6324	0.6835	0.7668	0.5405	0.6966	0.8200	0.7257	0.7640	0.7232
0.0863	93.3333	560	0.5737	0.6348	0.7544	0.7607	nan	0.7237	0.6606	0.7705	0.7775	0.6186	0.7401	0.8493	0.8438	0.7727	0.7868	0.0	0.7032	0.5988	0.7055	0.7366	0.5337	0.7141	0.8136	0.7060	0.7485	0.7230
0.072	96.6667	580	0.5544	0.6400	0.7616	0.7691	nan	0.7153	0.6546	0.7755	0.7964	0.6272	0.7607	0.8605	0.8571	0.7832	0.7854	0.0	0.6956	0.5972	0.7029	0.7550	0.5402	0.7088	0.8237	0.7215	0.7640	0.7309
0.1014	100.0	600	0.5610	0.6387	0.7597	0.7684	nan	0.7170	0.6456	0.7788	0.7965	0.6187	0.7553	0.8524	0.8602	0.7899	0.7831	0.0	0.6979	0.5897	0.7036	0.7569	0.5348	0.7058	0.8192	0.7187	0.7689	0.7295

Framework versions

• Transformers 4.52.2
• Pytorch 2.6.0+cu124
• Datasets 2.16.1
• Tokenizers 0.21.1