segformer-b5-finetuned-morphpadver1-hgo-coord-v6

This model is a fine-tuned version of nvidia/mit-b5 on the NICOPOI-9/morphpad_coord_hgo_512_4class_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0083
• Mean Iou: 0.9982
• Mean Accuracy: 0.9991
• Overall Accuracy: 0.9991
• Accuracy 0-0: 0.9993
• Accuracy 0-90: 0.9991
• Accuracy 90-0: 0.9996
• Accuracy 90-90: 0.9984
• Iou 0-0: 0.9988
• Iou 0-90: 0.9981
• Iou 90-0: 0.9979
• Iou 90-90: 0.9981

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: 1
• eval_batch_size: 1
• 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy 0-0	Accuracy 0-90	Accuracy 90-0	Accuracy 90-90	Iou 0-0	Iou 0-90	Iou 90-0	Iou 90-90
0.9111	2.6525	4000	0.8614	0.4113	0.5804	0.5812	0.6876	0.3429	0.6048	0.6862	0.5164	0.2855	0.3396	0.5035
0.4003	5.3050	8000	0.4114	0.6920	0.8177	0.8178	0.8421	0.7507	0.7478	0.9300	0.7679	0.6480	0.6597	0.6924
0.2246	7.9576	12000	0.2023	0.8443	0.9155	0.9155	0.9119	0.8979	0.8940	0.9583	0.8755	0.8355	0.8336	0.8327
0.1235	10.6101	16000	0.1268	0.9097	0.9527	0.9528	0.9612	0.9442	0.9450	0.9605	0.9223	0.9063	0.8991	0.9112
0.1012	13.2626	20000	0.0789	0.9445	0.9715	0.9715	0.9731	0.9707	0.9771	0.9650	0.9520	0.9426	0.9391	0.9445
0.0473	15.9151	24000	0.0582	0.9606	0.9799	0.9799	0.9769	0.9807	0.9832	0.9789	0.9615	0.9573	0.9588	0.9648
0.0258	18.5676	28000	0.0353	0.9830	0.9914	0.9914	0.9908	0.9915	0.9927	0.9906	0.9837	0.9824	0.9807	0.9850
0.046	21.2202	32000	0.0361	0.9839	0.9919	0.9919	0.9904	0.9916	0.9934	0.9922	0.9834	0.9832	0.9824	0.9866
0.0169	23.8727	36000	0.0262	0.9874	0.9937	0.9937	0.9937	0.9932	0.9935	0.9943	0.9883	0.9865	0.9871	0.9878
0.0127	26.5252	40000	0.0166	0.9926	0.9963	0.9963	0.9961	0.9957	0.9965	0.9968	0.9933	0.9918	0.9919	0.9934
0.0249	29.1777	44000	0.0222	0.9924	0.9962	0.9962	0.9931	0.9984	0.9962	0.9972	0.9913	0.9921	0.9919	0.9945
0.007	31.8302	48000	0.0114	0.9960	0.9980	0.9980	0.9979	0.9978	0.9981	0.9982	0.9960	0.9960	0.9956	0.9963
0.0061	34.4828	52000	0.0123	0.9966	0.9983	0.9983	0.9983	0.9981	0.9990	0.9978	0.9974	0.9963	0.9960	0.9965
0.0073	37.1353	56000	0.0125	0.9965	0.9983	0.9982	0.9977	0.9986	0.9985	0.9982	0.9968	0.9966	0.9961	0.9965
0.0053	39.7878	60000	0.0111	0.9974	0.9987	0.9987	0.9989	0.9985	0.9982	0.9993	0.9979	0.9974	0.9969	0.9975
0.0041	42.4403	64000	0.0125	0.9979	0.9989	0.9989	0.9988	0.9992	0.9991	0.9987	0.9979	0.9981	0.9976	0.9980
0.0037	45.0928	68000	0.0088	0.9980	0.9990	0.9990	0.9992	0.9991	0.9992	0.9985	0.9986	0.9980	0.9976	0.9979
0.0082	47.7454	72000	0.0083	0.9982	0.9991	0.9991	0.9993	0.9991	0.9996	0.9984	0.9988	0.9981	0.9979	0.9981

Framework versions

• Transformers 4.48.3
• Pytorch 2.1.0
• Datasets 3.2.0
• Tokenizers 0.21.0