segformer-b5-finetuned-ce-head-batch1

This model is a fine-tuned version of nvidia/mit-b5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0440
• Mean Iou: 0.7859
• Mean Accuracy: 0.8788
• Overall Accuracy: 0.9823
• Accuracy Bg: 0.9897
• Accuracy Head: 0.7680
• Iou Bg: 0.9819
• Iou Head: 0.5899

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: 5e-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: 80

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Bg	Accuracy Head	Iou Bg	Iou Head
0.0884	2.9412	100	0.1053	0.6562	0.7059	0.9673	0.9907	0.4211	0.9667	0.3456
0.1706	5.8824	200	0.0874	0.6241	0.6473	0.9660	0.9976	0.2969	0.9655	0.2828
0.0701	8.8235	300	0.0642	0.7050	0.7350	0.9749	0.9963	0.4736	0.9744	0.4357
0.0395	11.7647	400	0.0686	0.7212	0.7506	0.9742	0.9964	0.5048	0.9736	0.4688
0.0363	14.7059	500	0.0587	0.7770	0.8551	0.9765	0.9884	0.7217	0.9758	0.5783
0.1636	17.6471	600	0.0474	0.8091	0.9205	0.9800	0.9853	0.8556	0.9792	0.6389
0.0794	20.5882	700	0.0484	0.8021	0.8651	0.9799	0.9914	0.7389	0.9792	0.6251
0.0242	23.5294	800	0.0398	0.8238	0.8610	0.9845	0.9959	0.7261	0.9840	0.6635
0.0545	26.4706	900	0.0363	0.8395	0.8948	0.9855	0.9937	0.7960	0.9850	0.6941
0.0135	29.4118	1000	0.0381	0.8070	0.8337	0.9837	0.9973	0.6701	0.9833	0.6308
0.0212	32.3529	1100	0.0348	0.8396	0.8882	0.9856	0.9945	0.7819	0.9852	0.6939
0.0131	35.2941	1200	0.0405	0.8257	0.8536	0.9848	0.9971	0.7100	0.9844	0.6671
0.0212	38.2353	1300	0.0397	0.8341	0.9150	0.9841	0.9903	0.8398	0.9835	0.6848
0.0403	41.1765	1400	0.0335	0.8373	0.8855	0.9856	0.9946	0.7763	0.9851	0.6894
0.0611	44.1176	1500	0.0383	0.8352	0.8728	0.9847	0.9957	0.7498	0.9842	0.6863
0.0338	47.0588	1600	0.0305	0.8581	0.9177	0.9871	0.9933	0.8420	0.9866	0.7296
0.0286	50.0	1700	0.0462	0.8070	0.8293	0.9825	0.9978	0.6608	0.9819	0.6321
0.0459	52.9412	1800	0.0302	0.8634	0.9143	0.9878	0.9945	0.8341	0.9874	0.7393
0.0473	55.8824	1900	0.0389	0.8210	0.8524	0.9846	0.9967	0.7082	0.9842	0.6578
0.0059	58.8235	2000	0.0331	0.8454	0.8753	0.9868	0.9970	0.7536	0.9864	0.7044
0.0432	61.7647	2100	0.0412	0.8239	0.8554	0.9848	0.9967	0.7141	0.9843	0.6636
0.0153	64.7059	2200	0.0322	0.8465	0.8869	0.9866	0.9956	0.7781	0.9862	0.7068
0.0181	67.6471	2300	0.0335	0.8341	0.8716	0.9856	0.9959	0.7473	0.9852	0.6830
0.039	70.5882	2400	0.0347	0.8340	0.8794	0.9853	0.9949	0.7638	0.9848	0.6831
0.0212	73.5294	2500	0.0425	0.8014	0.8340	0.9816	0.9962	0.6718	0.9811	0.6218
0.0113	76.4706	2600	0.0318	0.8412	0.8837	0.9861	0.9953	0.7720	0.9856	0.6968
0.0642	79.4118	2700	0.0415	0.8153	0.8450	0.9831	0.9967	0.6932	0.9825	0.6481

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1
• Datasets 3.1.0
• Tokenizers 0.20.3