segformer-b0-finetuned-morphpadver1-hgo-30-coord-v3_120epochs

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: 0.3128
• Mean Iou: 0.7857
• Mean Accuracy: 0.8800
• Overall Accuracy: 0.8799
• Accuracy 0-0: 0.8782
• Accuracy 0-90: 0.8842
• Accuracy 90-0: 0.8809
• Accuracy 90-90: 0.8765
• Iou 0-0: 0.7869
• Iou 0-90: 0.7762
• Iou 90-0: 0.7880
• Iou 90-90: 0.7916

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: 120

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
1.188	4.2105	4000	1.2023	0.2499	0.3996	0.3995	0.3256	0.4999	0.4603	0.3124	0.2510	0.2542	0.2429	0.2514
0.9829	8.4211	8000	0.9446	0.3676	0.5326	0.5326	0.4568	0.5068	0.7157	0.4508	0.3872	0.3492	0.3434	0.3907
0.7539	12.6316	12000	0.7804	0.4475	0.6163	0.6168	0.5690	0.6618	0.5314	0.7030	0.4806	0.4244	0.4335	0.4513
0.969	16.8421	16000	0.6554	0.5152	0.6775	0.6780	0.6172	0.7013	0.6543	0.7372	0.5533	0.4849	0.5090	0.5136
0.5649	21.0526	20000	0.5879	0.5597	0.7156	0.7157	0.6964	0.7740	0.6842	0.7078	0.5869	0.5181	0.5555	0.5782
0.5983	25.2632	24000	0.5229	0.5963	0.7464	0.7460	0.8236	0.7139	0.7132	0.7348	0.5647	0.6058	0.6038	0.6110
0.5077	29.4737	28000	0.4971	0.6133	0.7587	0.7585	0.7767	0.7577	0.7964	0.7042	0.6236	0.6053	0.5724	0.6519
0.456	33.6842	32000	0.4880	0.6346	0.7763	0.7764	0.7792	0.7575	0.7675	0.8011	0.6434	0.6257	0.6376	0.6317
0.4278	37.8947	36000	0.4139	0.6688	0.8015	0.8014	0.8149	0.7962	0.8101	0.7849	0.6663	0.6599	0.6611	0.6880
0.4974	42.1053	40000	0.3921	0.6863	0.8132	0.8132	0.7997	0.8370	0.8165	0.7996	0.7065	0.6548	0.6794	0.7046
0.4364	46.3158	44000	0.3697	0.7023	0.8244	0.8247	0.7941	0.8293	0.8225	0.8520	0.7248	0.6884	0.6982	0.6978
0.3254	50.5263	48000	0.3521	0.7152	0.8340	0.8338	0.8529	0.8268	0.8315	0.8247	0.7016	0.7133	0.7162	0.7295
0.3139	54.7368	52000	0.3471	0.7224	0.8386	0.8386	0.8343	0.8526	0.8305	0.8371	0.7209	0.7032	0.7306	0.7347
0.3209	58.9474	56000	0.3253	0.7359	0.8479	0.8479	0.8565	0.8363	0.8525	0.8463	0.7340	0.7348	0.7294	0.7455
0.2815	63.1579	60000	0.3234	0.7421	0.8516	0.8516	0.8468	0.8707	0.8431	0.8459	0.7466	0.7148	0.7463	0.7608
0.3002	67.3684	64000	0.3132	0.7520	0.8584	0.8584	0.8545	0.8623	0.8503	0.8663	0.7544	0.7411	0.7560	0.7566
0.2874	71.5789	68000	0.3068	0.7571	0.8615	0.8615	0.8582	0.8814	0.8540	0.8524	0.7628	0.7325	0.7639	0.7693
1.0781	75.7895	72000	0.3185	0.7524	0.8588	0.8586	0.8755	0.8528	0.8649	0.8419	0.7446	0.7461	0.7503	0.7686
0.2688	80.0	76000	0.2993	0.7663	0.8676	0.8676	0.8688	0.8677	0.8639	0.8702	0.7693	0.7553	0.7674	0.7730
0.2566	84.2105	80000	0.2962	0.7696	0.8698	0.8698	0.8669	0.8687	0.8673	0.8761	0.7729	0.7574	0.7739	0.7744
0.2556	88.4211	84000	0.2985	0.7754	0.8735	0.8735	0.8735	0.8679	0.8799	0.8726	0.7767	0.7676	0.7734	0.7839
0.2402	92.6316	88000	0.2976	0.7786	0.8755	0.8755	0.8786	0.8751	0.8694	0.8790	0.7766	0.7711	0.7810	0.7858
0.286	96.8421	92000	0.2998	0.7803	0.8766	0.8766	0.8733	0.8750	0.8828	0.8752	0.7833	0.7733	0.7770	0.7876
0.1853	101.0526	96000	0.2987	0.7843	0.8791	0.8791	0.8810	0.8816	0.8813	0.8726	0.7811	0.7754	0.7870	0.7935
0.2401	105.2632	100000	0.3093	0.7819	0.8776	0.8776	0.8761	0.8820	0.8791	0.8732	0.7818	0.7701	0.7864	0.7893
0.2546	109.4737	104000	0.3095	0.7846	0.8793	0.8793	0.8819	0.8829	0.8756	0.8766	0.7850	0.7731	0.7891	0.7912
0.3595	113.6842	108000	0.3096	0.7857	0.8800	0.8800	0.8777	0.8820	0.8795	0.8806	0.7858	0.7766	0.7892	0.7912
0.2484	117.8947	112000	0.3128	0.7857	0.8800	0.8799	0.8782	0.8842	0.8809	0.8765	0.7869	0.7762	0.7880	0.7916

Framework versions

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