segformer-b0-finetuned-morphpadver1-hgo-30-coord-v3_60epochs

This model is a fine-tuned version of nvidia/mit-b3 on the NICOPOI-9/morphpad_coord_hgo_30_30_512_4class dataset. It achieves the following results on the evaluation set:

• Loss: 0.5626
• Mean Iou: 0.5820
• Mean Accuracy: 0.7358
• Overall Accuracy: 0.7358
• Accuracy 0-0: 0.7456
• Accuracy 0-90: 0.7128
• Accuracy 90-0: 0.7363
• Accuracy 90-90: 0.7484
• Iou 0-0: 0.5840
• Iou 0-90: 0.5781
• Iou 90-0: 0.5720
• Iou 90-90: 0.5939

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
1.2478	4.2105	4000	1.2564	0.2012	0.3564	0.3563	0.2794	0.7320	0.1626	0.2516	0.2015	0.2645	0.1424	0.1964
1.1864	8.4211	8000	1.0945	0.2822	0.4420	0.4430	0.3826	0.4025	0.3519	0.6312	0.2902	0.2692	0.2660	0.3036
0.9632	12.6316	12000	0.9682	0.3432	0.5103	0.5103	0.4745	0.4817	0.6326	0.4526	0.3457	0.3355	0.3377	0.3539
1.0223	16.8421	16000	0.8653	0.4020	0.5689	0.5690	0.4846	0.7162	0.5767	0.4982	0.4109	0.3743	0.3890	0.4336
0.7388	21.0526	20000	0.7888	0.4382	0.6064	0.6068	0.5402	0.6197	0.6163	0.6494	0.4767	0.4090	0.4268	0.4403
0.7634	25.2632	24000	0.7226	0.4711	0.6404	0.6406	0.6547	0.6184	0.5925	0.6962	0.4872	0.4634	0.4606	0.4733
0.6536	29.4737	28000	0.6801	0.4993	0.6654	0.6657	0.6463	0.6443	0.6653	0.7058	0.5182	0.4909	0.4806	0.5074
0.6216	33.6842	32000	0.6512	0.5192	0.6821	0.6826	0.6793	0.6185	0.6460	0.7848	0.5362	0.5204	0.5184	0.5019
0.6402	37.8947	36000	0.6295	0.5309	0.6932	0.6931	0.7050	0.7227	0.6512	0.6938	0.5298	0.5108	0.5348	0.5482
0.7389	42.1053	40000	0.6110	0.5475	0.7076	0.7077	0.7126	0.6793	0.7010	0.7374	0.5522	0.5449	0.5321	0.5610
0.6753	46.3158	44000	0.5858	0.5631	0.7203	0.7202	0.7338	0.6868	0.7393	0.7212	0.5700	0.5556	0.5437	0.5831
0.4944	50.5263	48000	0.5762	0.5711	0.7264	0.7264	0.7266	0.6984	0.7537	0.7268	0.5827	0.5703	0.5430	0.5885
0.4953	54.7368	52000	0.5676	0.5804	0.7337	0.7336	0.7532	0.6790	0.7716	0.7310	0.5759	0.5939	0.5535	0.5983
0.4828	58.9474	56000	0.5626	0.5820	0.7358	0.7358	0.7456	0.7128	0.7363	0.7484	0.5840	0.5781	0.5720	0.5939

Framework versions

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