50d_seg_model_20250507

This model is a fine-tuned version of nvidia/mit-b0 on the TommyClas/50d_seg_20250505 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8496
• Mean Iou: 0.4618
• Mean Accuracy: 0.6974
• Overall Accuracy: 0.7722
• Accuracy 背景: nan
• Accuracy 孔隙: 0.7921
• Accuracy Ld c-s-h: 0.7918
• Accuracy Hd c-s-h: 0.3186
• Accuracy 未水化水泥颗粒: 0.8868
• Iou 背景: 0.0
• Iou 孔隙: 0.6924
• Iou Ld c-s-h: 0.5935
• Iou Hd c-s-h: 0.2372
• Iou 未水化水泥颗粒: 0.7862

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: 0.01
• train_batch_size: 8
• eval_batch_size: 8
• seed: 1337
• 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: polynomial
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Accuracy Hd c-s-h	Accuracy Ld c-s-h	Accuracy 孔隙	Accuracy 未水化水泥颗粒	Accuracy 背景	Iou Hd c-s-h	Iou Ld c-s-h	Iou 孔隙	Iou 未水化水泥颗粒	Iou 背景	Validation Loss	Mean Accuracy	Mean Iou	Overall Accuracy
1.0019	1.0	250	0.0	0.0650	0.7441	0.4815	nan	0.0	0.0544	0.3893	0.4701	0.0	4.4151	0.3226	0.1828	0.3786
0.4554	2.0	500	0.0000	0.6842	0.7487	0.8724	nan	0.0000	0.5197	0.5813	0.7626	0.0	1.2340	0.5763	0.3727	0.6884
0.4529	3.0	750	0.0000	0.7384	0.7863	0.8553	nan	0.0000	0.5339	0.6291	0.7592	0.0	0.9681	0.5950	0.3844	0.7172
0.4248	4.0	1000	0.2327	0.5245	0.8565	0.8152	nan	0.1599	0.4278	0.5931	0.7471	0.0	1.0964	0.6072	0.3856	0.6744
0.4194	5.0	1250	0.2258	0.7465	0.4263	0.9194	nan	0.1022	0.4743	0.4199	0.7475	0.0	0.9903	0.5795	0.3488	0.6308
0.4183	6.0	1500	0.5113	0.4379	0.9017	0.7679	nan	0.1996	0.3818	0.6551	0.7309	0.0	0.9263	0.6547	0.3935	0.6717
0.4178	7.0	1750	0.2577	0.5691	0.9078	0.8036	nan	0.1468	0.4584	0.6702	0.7487	0.0	0.8280	0.6346	0.4048	0.7077
0.4144	8.0	2000	0.1812	0.6915	0.6035	0.9596	nan	0.1060	0.4844	0.5706	0.6832	0.0	0.9423	0.6090	0.3688	0.6757
0.4148	9.0	2250	0.3482	0.1554	0.9229	0.5755	nan	0.1635	0.1371	0.4818	0.5680	0.0	1.7458	0.5005	0.2701	0.5215
0.4051	10.0	2500	0.0512	0.7194	0.8609	0.7365	nan	0.0456	0.5033	0.6617	0.6965	0.0	0.8936	0.5920	0.3814	0.7146
0.4027	11.0	2750	0.1803	0.5811	0.9251	0.8497	nan	0.1368	0.4711	0.6614	0.7522	0.0	0.8490	0.6340	0.4043	0.7212
0.4	12.0	3000	0.0871	0.7090	0.8824	0.8546	nan	0.0746	0.5374	0.6926	0.7589	0.0	0.7884	0.6333	0.4127	0.7463
0.3966	13.0	3250	0.3821	0.7329	0.8081	0.7223	nan	0.2117	0.5417	0.6819	0.6861	0.0	0.7998	0.6614	0.4243	0.7265
0.3809	14.0	3500	0.1311	0.6534	0.9270	0.8173	nan	0.1055	0.5109	0.6812	0.7604	0.0	0.8045	0.6322	0.4116	0.7370
0.3769	15.0	3750	0.2447	0.7358	0.3770	0.9680	nan	0.1478	0.4557	0.3679	0.6794	0.0	1.0931	0.5814	0.3301	0.6222
0.3742	16.0	4000	0.0718	0.6090	0.9529	0.8171	nan	0.0647	0.4802	0.6614	0.7576	0.0	0.8481	0.6127	0.3928	0.7246
0.3723	17.0	4250	0.1708	0.7345	0.8772	0.8962	nan	0.1393	0.5764	0.7142	0.7782	0.0	0.6791	0.6697	0.4416	0.7698
0.369	18.0	4500	0.0727	0.7378	0.9273	0.7286	nan	0.0679	0.5349	0.7077	0.7059	0.0	0.7373	0.6166	0.4033	0.7440
0.3812	19.0	4750	0.1705	0.8129	0.8411	0.8514	nan	0.1543	0.5980	0.7109	0.7786	0.0	0.6597	0.6690	0.4484	0.7764
0.3686	20.0	5000	0.1371	0.7812	0.8773	0.8351	nan	0.1248	0.5836	0.7143	0.7709	0.0	0.7417	0.6577	0.4387	0.7709
0.3675	21.0	5250	0.2222	0.7268	0.9023	0.8715	nan	0.1882	0.5762	0.7135	0.7853	0.0	0.7173	0.6807	0.4526	0.7746
0.3647	22.0	5500	0.2299	0.7688	0.8603	0.8955	nan	0.1904	0.5974	0.7207	0.7851	0.0	0.7049	0.6886	0.4587	0.7813
0.3629	23.0	5750	0.2192	0.6911	0.9342	0.8092	nan	0.1833	0.5437	0.7023	0.7632	0.0	0.8170	0.6634	0.4385	0.7589
0.3623	24.0	6000	0.2782	0.7554	0.8817	0.8655	nan	0.2243	0.5923	0.7212	0.7869	0.0	0.7217	0.6952	0.4649	0.7814
0.3609	25.0	6250	0.2384	0.7263	0.9150	0.8090	nan	0.1964	0.5627	0.7133	0.7635	0.0	0.8099	0.6722	0.4472	0.7667
0.3619	26.0	6500	0.2477	0.8017	0.8433	0.8570	nan	0.2032	0.6010	0.7189	0.7839	0.0	0.7394	0.6874	0.4614	0.7808
0.3603	27.0	6750	0.2353	0.8035	0.8323	0.8729	nan	0.1923	0.6014	0.7161	0.7879	0.0	0.7378	0.6860	0.4595	0.7801
0.3591	28.0	7000	0.2726	0.7587	0.8749	0.8612	nan	0.2141	0.5885	0.7218	0.7867	0.0	0.7468	0.6918	0.4622	0.7789
0.3584	29.0	7250	0.7495	0.4603	0.6958	0.7760	nan	0.7940	0.7926	0.2814	0.9150	0.0	0.7036	0.5987	0.2133	0.7861
0.3564	30.0	7500	0.7896	0.4520	0.6803	0.7733	nan	0.7788	0.8223	0.2201	0.9001	0.0	0.6955	0.6001	0.1781	0.7865
0.3567	31.0	7750	0.7667	0.4590	0.6863	0.7810	nan	0.8533	0.7825	0.2269	0.8827	0.0	0.7195	0.5977	0.1887	0.7891
0.3551	32.0	8000	0.7786	0.4616	0.6945	0.7776	nan	0.7922	0.8068	0.2755	0.9036	0.0	0.7005	0.6029	0.2179	0.7866
0.3539	33.0	8250	0.7576	0.4644	0.6960	0.7801	nan	0.8306	0.7869	0.2804	0.8862	0.0	0.7122	0.5988	0.2222	0.7889
0.354	34.0	8500	0.7926	0.4632	0.6966	0.7774	nan	0.7967	0.8027	0.2899	0.8969	0.0	0.7016	0.6007	0.2255	0.7881
0.3519	35.0	8750	0.7779	0.4626	0.6919	0.7771	nan	0.8519	0.7782	0.2869	0.8505	0.0	0.7137	0.5914	0.2252	0.7825
0.3511	36.0	9000	0.8104	0.4638	0.6987	0.7753	nan	0.7967	0.7967	0.3138	0.8878	0.0	0.6982	0.5980	0.2358	0.7873
0.3505	37.0	9250	0.8228	0.4613	0.6954	0.7735	nan	0.7830	0.8048	0.3016	0.8922	0.0	0.6916	0.5975	0.2307	0.7868
0.3499	38.0	9500	0.8103	0.4617	0.6939	0.7752	nan	0.8289	0.7767	0.2915	0.8786	0.0	0.7023	0.5905	0.2290	0.7869
0.3493	39.0	9750	0.8413	0.4616	0.6951	0.7733	nan	0.8075	0.7885	0.3065	0.8778	0.0	0.6970	0.5925	0.2325	0.7863
0.3481	40.0	10000	0.8496	0.4618	0.6974	0.7722	nan	0.7921	0.7918	0.3186	0.8868	0.0	0.6924	0.5935	0.2372	0.7862

Framework versions

• Transformers 4.52.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1