segformer-b5-finetuned-morphpadver1-hgo-coord-v9_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: 0.9212
• Mean Iou: 0.6264
• Mean Accuracy: 0.7657
• Overall Accuracy: 0.7693
• Accuracy 0-0: 0.7303
• Accuracy 0-90: 0.8137
• Accuracy 90-0: 0.7916
• Accuracy 90-90: 0.7271
• Iou 0-0: 0.6366
• Iou 0-90: 0.6229
• Iou 90-0: 0.6132
• Iou 90-90: 0.6329

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
1.3571	1.3638	4000	1.3517	0.1822	0.3201	0.3311	0.2216	0.2676	0.6249	0.1665	0.1575	0.1831	0.2565	0.1315
0.8199	2.7276	8000	1.2275	0.2731	0.4274	0.4361	0.3563	0.5122	0.5256	0.3155	0.2523	0.3082	0.2937	0.2381
0.8824	4.0914	12000	1.1421	0.3520	0.5198	0.5231	0.4839	0.5162	0.5954	0.4839	0.3342	0.3733	0.3710	0.3297
0.5435	5.4552	16000	0.9993	0.4242	0.5921	0.5979	0.5242	0.6641	0.6381	0.5420	0.4136	0.4409	0.4348	0.4076
0.8088	6.8190	20000	1.0559	0.4473	0.6166	0.6183	0.5671	0.5950	0.6749	0.6296	0.4524	0.4525	0.4513	0.4331
0.3228	8.1827	24000	0.9718	0.4925	0.6572	0.6604	0.5965	0.6694	0.7118	0.6511	0.4794	0.5029	0.4892	0.4985
0.8418	9.5465	28000	0.9748	0.5147	0.6735	0.6808	0.6234	0.7941	0.6989	0.5776	0.5228	0.5218	0.5217	0.4925
0.4066	10.9103	32000	0.9678	0.5360	0.6956	0.6985	0.6499	0.7135	0.7388	0.6803	0.5274	0.5461	0.5374	0.5330
0.3456	12.2741	36000	0.8965	0.5680	0.7221	0.7245	0.6491	0.7611	0.7252	0.7532	0.5661	0.5709	0.5625	0.5725
0.3544	13.6379	40000	0.8759	0.5800	0.7301	0.7343	0.7005	0.8018	0.7436	0.6744	0.5869	0.5831	0.5780	0.5721
0.3027	15.0017	44000	0.8860	0.5966	0.7437	0.7471	0.6909	0.7757	0.7824	0.7257	0.6008	0.5977	0.5931	0.5947
0.1839	16.3655	48000	0.9557	0.6063	0.7507	0.7537	0.7106	0.7862	0.7744	0.7317	0.6161	0.6070	0.5849	0.6170
0.1924	17.7293	52000	0.8912	0.6285	0.7682	0.7711	0.7340	0.8063	0.7894	0.7432	0.6382	0.6315	0.6125	0.6315
0.2531	19.0931	56000	0.9212	0.6264	0.7657	0.7693	0.7303	0.8137	0.7916	0.7271	0.6366	0.6229	0.6132	0.6329

Framework versions

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