segformer-b5-finetuned-morphpadver1-hgo-coord-v8_mix_resample_20epochs

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: 0.1662
• Mean Iou: 0.9145
• Mean Accuracy: 0.9539
• Overall Accuracy: 0.9551
• Accuracy 0-0: 0.9434
• Accuracy 0-90: 0.9670
• Accuracy 90-0: 0.9662
• Accuracy 90-90: 0.9390
• Iou 0-0: 0.9189
• Iou 0-90: 0.9107
• Iou 90-0: 0.9106
• Iou 90-90: 0.9181

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: 20

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
1.3433	1.3638	4000	1.3061	0.2130	0.3543	0.3666	0.2175	0.5568	0.3968	0.2460	0.1696	0.2670	0.2344	0.1813
0.6413	2.7276	8000	0.9887	0.3847	0.5482	0.5575	0.4684	0.7205	0.5574	0.4468	0.3795	0.3902	0.3956	0.3736
0.5943	4.0914	12000	0.7461	0.5269	0.6874	0.6893	0.6820	0.6888	0.7283	0.6505	0.5493	0.5185	0.5213	0.5186
0.3558	5.4552	16000	0.6020	0.6131	0.7528	0.7586	0.7312	0.8810	0.7401	0.6588	0.6330	0.5916	0.6263	0.6013
0.4599	6.8190	20000	0.4441	0.7185	0.8303	0.8347	0.8221	0.9146	0.8356	0.7488	0.7563	0.6939	0.7257	0.6979
0.2557	8.1827	24000	0.3843	0.7536	0.8536	0.8587	0.7902	0.9352	0.8732	0.8158	0.7588	0.7358	0.7633	0.7563
0.5117	9.5465	28000	0.3007	0.8088	0.8914	0.8931	0.8711	0.9330	0.8824	0.8789	0.8237	0.7859	0.8062	0.8196
0.1445	10.9103	32000	0.2623	0.8378	0.9114	0.9111	0.9184	0.9072	0.9090	0.9110	0.8475	0.8254	0.8302	0.8480
0.1051	12.2741	36000	0.2383	0.8477	0.9174	0.9171	0.9327	0.9096	0.9203	0.9070	0.8655	0.8462	0.8332	0.8459
0.1346	13.6379	40000	0.2149	0.8673	0.9272	0.9286	0.9180	0.9429	0.9402	0.9077	0.8763	0.8609	0.8624	0.8697
0.1045	15.0017	44000	0.1777	0.8982	0.9464	0.9462	0.9465	0.9345	0.9540	0.9507	0.9009	0.8964	0.8925	0.9030
0.0383	16.3655	48000	0.1812	0.9011	0.9466	0.9478	0.9454	0.9574	0.9603	0.9235	0.9072	0.8996	0.8965	0.9010
0.0416	17.7293	52000	0.1494	0.9167	0.9556	0.9564	0.9524	0.9651	0.9626	0.9422	0.9215	0.9158	0.9114	0.9181
0.0732	19.0931	56000	0.1662	0.9145	0.9539	0.9551	0.9434	0.9670	0.9662	0.9390	0.9189	0.9107	0.9106	0.9181

Framework versions

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