segformer-b5-finetuned-ade20k-morphpadver1-hgo-coord-v9_mix_resample_40epochs_distortion

This model is a fine-tuned version of nvidia/segformer-b5-finetuned-ade-640-640 on the NICOPOI-9/morphpad_coord_hgo_512_4class_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0248
• Mean Iou: 0.9870
• Mean Accuracy: 0.9930
• Overall Accuracy: 0.9935
• Accuracy 0-0: 0.9819
• Accuracy 0-90: 0.9979
• Accuracy 90-0: 0.9981
• Accuracy 90-90: 0.9941
• Iou 0-0: 0.9810
• Iou 0-90: 0.9907
• Iou 90-0: 0.9849
• Iou 90-90: 0.9915

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: 40

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
1.4019	1.3638	4000	1.1504	0.3001	0.4634	0.4755	0.2605	0.6250	0.5404	0.4276	0.2154	0.3647	0.3500	0.2704
0.4906	2.7276	8000	0.7690	0.5119	0.6703	0.6837	0.5029	0.7631	0.8583	0.5568	0.4600	0.5657	0.5274	0.4947
0.4358	4.0914	12000	0.5382	0.6685	0.8006	0.8017	0.7921	0.7457	0.8873	0.7774	0.6784	0.6729	0.6762	0.6466
0.5753	5.4552	16000	0.3940	0.7475	0.8549	0.8567	0.7635	0.8536	0.8910	0.9112	0.7177	0.7860	0.7475	0.7390
0.152	6.8190	20000	0.1956	0.8760	0.9319	0.9343	0.8673	0.9464	0.9646	0.9492	0.8509	0.8754	0.8824	0.8955
0.2229	8.1827	24000	0.1526	0.9089	0.9513	0.9524	0.9276	0.9543	0.9693	0.9542	0.9048	0.9166	0.9042	0.9099
0.6554	9.5465	28000	0.1451	0.9003	0.9451	0.9480	0.8843	0.9647	0.9838	0.9477	0.8695	0.9214	0.8916	0.9188
0.9251	10.9103	32000	0.1035	0.9346	0.9648	0.9666	0.9143	0.9806	0.9848	0.9794	0.9099	0.9443	0.9359	0.9481
1.022	12.2741	36000	0.0706	0.9594	0.9790	0.9793	0.9668	0.9776	0.9868	0.9845	0.9588	0.9615	0.9580	0.9591
0.0331	13.6379	40000	0.0763	0.9503	0.9738	0.9746	0.9621	0.9848	0.9787	0.9695	0.9402	0.9626	0.9399	0.9587
0.0458	15.0017	44000	0.0649	0.9570	0.9770	0.9781	0.9538	0.9812	0.9948	0.9784	0.9475	0.9663	0.9491	0.9653
0.0271	16.3655	48000	0.0504	0.9719	0.9850	0.9857	0.9765	0.9916	0.9929	0.9790	0.9726	0.9763	0.9652	0.9736
0.0763	17.7293	52000	0.0452	0.9689	0.9837	0.9844	0.9641	0.9880	0.9922	0.9906	0.9565	0.9771	0.9668	0.9753
1.054	19.0931	56000	0.0531	0.9746	0.9864	0.9871	0.9756	0.9935	0.9937	0.9828	0.9733	0.9783	0.9701	0.9766
0.9113	20.4569	60000	0.0400	0.9797	0.9894	0.9898	0.9802	0.9936	0.9931	0.9906	0.9747	0.9846	0.9747	0.9850
8.3984	21.8207	64000	0.0360	0.9844	0.9918	0.9921	0.9869	0.9958	0.9941	0.9903	0.9836	0.9861	0.9809	0.9870
0.0613	23.1845	68000	0.0380	0.9812	0.9900	0.9905	0.9832	0.9961	0.9941	0.9864	0.9798	0.9843	0.9768	0.9838
1.015	24.5482	72000	0.0361	0.9836	0.9917	0.9917	0.9892	0.9933	0.9916	0.9925	0.9821	0.9855	0.9809	0.9861
0.0142	25.9120	76000	0.0371	0.9749	0.9868	0.9873	0.9760	0.9925	0.9911	0.9878	0.9692	0.9801	0.9712	0.9790
0.0477	27.2758	80000	0.0420	0.9822	0.9907	0.9911	0.9794	0.9961	0.9921	0.9952	0.9729	0.9886	0.9772	0.9900
0.0562	28.6396	84000	0.0402	0.9775	0.9881	0.9888	0.9681	0.9957	0.9934	0.9953	0.9646	0.9818	0.9783	0.9854
0.9959	30.0034	88000	0.0294	0.9853	0.9925	0.9926	0.9902	0.9955	0.9901	0.9944	0.9809	0.9883	0.9820	0.9899
0.0019	31.3672	92000	0.0309	0.9834	0.9911	0.9918	0.9768	0.9984	0.9963	0.9929	0.9750	0.9871	0.9832	0.9885
0.3517	32.7310	96000	0.0243	0.9885	0.9940	0.9942	0.9896	0.9961	0.9965	0.9936	0.9868	0.9911	0.9854	0.9905
0.0101	34.0948	100000	0.0282	0.9863	0.9927	0.9931	0.9835	0.9962	0.9975	0.9936	0.9823	0.9885	0.9851	0.9893
0.0057	35.4586	104000	0.0221	0.9882	0.9938	0.9941	0.9850	0.9964	0.9980	0.9956	0.9842	0.9918	0.9861	0.9907
0.0099	36.8224	108000	0.0264	0.9889	0.9940	0.9944	0.9854	0.9976	0.9982	0.9949	0.9846	0.9925	0.9853	0.9930
0.0099	38.1862	112000	0.0229	0.9886	0.9939	0.9943	0.9860	0.9978	0.9979	0.9938	0.9850	0.9915	0.9862	0.9917
0.8071	39.5499	116000	0.0248	0.9870	0.9930	0.9935	0.9819	0.9979	0.9981	0.9941	0.9810	0.9907	0.9849	0.9915

Framework versions

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