README.md

metadata

library_name: transformers
license: other
base_model: nvidia/mit-b3
tags:
  - vision
  - image-segmentation
  - generated_from_trainer
model-index:
  - name: segformer-b0-finetuned-morphpadver1-hgo-coord-v3_1
    results: []

segformer-b0-finetuned-morphpadver1-hgo-coord-v3_1

This model is a fine-tuned version of nvidia/mit-b3 on the NICOPOI-9/morphpad_coord_hgo_512_4class_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0117
• Mean Iou: 0.9981
• Mean Accuracy: 0.9990
• Overall Accuracy: 0.9990
• Accuracy 0-0: 0.9995
• Accuracy 0-90: 0.9985
• Accuracy 90-0: 0.9988
• Accuracy 90-90: 0.9993
• Iou 0-0: 0.9991
• Iou 0-90: 0.9979
• Iou 90-0: 0.9976
• Iou 90-90: 0.9978

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: 60

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
0.903	2.6525	4000	0.8952	0.3916	0.5570	0.5567	0.5335	0.6125	0.4890	0.5929	0.4534	0.3418	0.3411	0.4299
0.6373	5.3050	8000	0.5078	0.6237	0.7643	0.7643	0.7676	0.8339	0.6741	0.7817	0.6758	0.5472	0.6022	0.6698
0.2851	7.9576	12000	0.2955	0.7612	0.8642	0.8642	0.8669	0.8687	0.8339	0.8874	0.7959	0.7358	0.7500	0.7631
0.2309	10.6101	16000	0.1305	0.9184	0.9574	0.9574	0.9575	0.9381	0.9648	0.9692	0.9333	0.9074	0.8991	0.9337
0.0907	13.2626	20000	0.1249	0.9267	0.9620	0.9620	0.9636	0.9541	0.9594	0.9708	0.9379	0.9205	0.9169	0.9316
0.3051	15.9151	24000	0.0529	0.9675	0.9835	0.9835	0.9842	0.9805	0.9839	0.9854	0.9712	0.9636	0.9626	0.9728
0.0659	18.5676	28000	0.0630	0.9670	0.9832	0.9833	0.9852	0.9747	0.9885	0.9846	0.9719	0.9642	0.9633	0.9687
0.0474	21.2202	32000	0.0454	0.9768	0.9882	0.9883	0.9910	0.9856	0.9865	0.9899	0.9783	0.9737	0.9747	0.9805
0.0449	23.8727	36000	0.0468	0.9795	0.9896	0.9896	0.9900	0.9812	0.9900	0.9973	0.9828	0.9743	0.9783	0.9824
0.0552	26.5252	40000	0.0266	0.9884	0.9942	0.9942	0.9949	0.9917	0.9947	0.9953	0.9888	0.9865	0.9866	0.9916
0.0541	29.1777	44000	0.0290	0.9908	0.9954	0.9954	0.9951	0.9951	0.9967	0.9946	0.9921	0.9897	0.9905	0.9909
0.0082	31.8302	48000	0.0421	0.9891	0.9945	0.9945	0.9940	0.9924	0.9951	0.9966	0.9908	0.9869	0.9884	0.9904
0.0061	34.4828	52000	0.0345	0.9923	0.9961	0.9961	0.9971	0.9941	0.9966	0.9966	0.9939	0.9912	0.9916	0.9922
0.0053	37.1353	56000	0.0256	0.9941	0.9970	0.9970	0.9976	0.9972	0.9966	0.9968	0.9957	0.9928	0.9929	0.9949
0.0045	39.7878	60000	0.0256	0.9937	0.9968	0.9968	0.9978	0.9959	0.9959	0.9978	0.9937	0.9927	0.9926	0.9957
0.0046	42.4403	64000	0.0171	0.9964	0.9982	0.9982	0.9983	0.9976	0.9987	0.9981	0.9972	0.9958	0.9955	0.9969
0.0032	45.0928	68000	0.0293	0.9957	0.9979	0.9979	0.9983	0.9969	0.9975	0.9988	0.9966	0.9950	0.9950	0.9964
0.003	47.7454	72000	0.0251	0.9964	0.9982	0.9982	0.9984	0.9973	0.9984	0.9987	0.9973	0.9952	0.9965	0.9966
0.0035	50.3979	76000	0.0245	0.9973	0.9986	0.9986	0.9993	0.9982	0.9983	0.9987	0.9982	0.9969	0.9963	0.9977
0.0025	53.0504	80000	0.0222	0.9972	0.9986	0.9986	0.9990	0.9980	0.9987	0.9986	0.9985	0.9965	0.9970	0.9968
0.0023	55.7029	84000	0.0104	0.9982	0.9991	0.9991	0.9994	0.9989	0.9987	0.9993	0.9988	0.9980	0.9975	0.9983
0.0022	58.3554	88000	0.0117	0.9981	0.9990	0.9990	0.9995	0.9985	0.9988	0.9993	0.9991	0.9979	0.9976	0.9978

Framework versions

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