segformer-b5-finetuned-morphpadver1-hgo-coord-v7_mix

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: 0.0252
• Mean Iou: 0.9929
• Mean Accuracy: 0.9964
• Overall Accuracy: 0.9964
• Accuracy 0-0: 0.9964
• Accuracy 0-90: 0.9977
• Accuracy 90-0: 0.9952
• Accuracy 90-90: 0.9962
• Iou 0-0: 0.9940
• Iou 0-90: 0.9938
• Iou 90-0: 0.9903
• Iou 90-90: 0.9936

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: 30

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
1.0592	1.3638	4000	1.0170	0.3508	0.5140	0.5289	0.3683	0.7216	0.6097	0.3563	0.3302	0.3825	0.3737	0.3169
0.721	2.7276	8000	0.5930	0.6031	0.7460	0.7502	0.6848	0.7925	0.7840	0.7227	0.6302	0.5834	0.5940	0.6049
0.29	4.0914	12000	0.3953	0.7256	0.8415	0.8408	0.8628	0.8237	0.8461	0.8335	0.7154	0.7324	0.7159	0.7387
0.2193	5.4552	16000	0.3058	0.7892	0.8811	0.8818	0.8586	0.8839	0.8933	0.8886	0.7896	0.7973	0.7726	0.7972
0.2548	6.8190	20000	0.2064	0.8644	0.9267	0.9268	0.9283	0.9250	0.9304	0.9232	0.8708	0.8602	0.8539	0.8726
0.1537	8.1827	24000	0.1766	0.8894	0.9406	0.9413	0.9321	0.9447	0.9511	0.9347	0.8805	0.8806	0.8949	0.9016
0.1259	9.5465	28000	0.1421	0.9240	0.9605	0.9602	0.9644	0.9561	0.9593	0.9621	0.9334	0.9180	0.9183	0.9265
0.0919	10.9103	32000	0.1213	0.9359	0.9673	0.9668	0.9708	0.9672	0.9563	0.9750	0.9298	0.9389	0.9293	0.9456
0.0416	12.2741	36000	0.0820	0.9569	0.9782	0.9778	0.9817	0.9709	0.9783	0.9818	0.9569	0.9530	0.9530	0.9649
0.0618	13.6379	40000	0.0742	0.9636	0.9815	0.9814	0.9845	0.9793	0.9811	0.9813	0.9600	0.9663	0.9590	0.9689
0.0553	15.0017	44000	0.0706	0.9699	0.9848	0.9847	0.9843	0.9836	0.9848	0.9863	0.9619	0.9708	0.9688	0.9781
0.0451	16.3655	48000	0.0789	0.9724	0.9863	0.9860	0.9918	0.9816	0.9850	0.9869	0.9671	0.9735	0.9706	0.9786
0.0123	17.7293	52000	0.0733	0.9746	0.9874	0.9871	0.9923	0.9834	0.9861	0.9876	0.9706	0.9735	0.9741	0.9803
0.0255	19.0931	56000	0.0400	0.9831	0.9916	0.9914	0.9919	0.9872	0.9927	0.9946	0.9829	0.9808	0.9820	0.9869
0.0124	20.4569	60000	0.0584	0.9830	0.9915	0.9914	0.9937	0.9947	0.9867	0.9908	0.9810	0.9845	0.9796	0.9870
20.9459	21.8207	64000	0.0300	0.9884	0.9942	0.9941	0.9963	0.9921	0.9939	0.9946	0.9914	0.9875	0.9873	0.9874
0.0036	23.1845	68000	0.0467	0.9836	0.9918	0.9917	0.9941	0.9850	0.9954	0.9928	0.9893	0.9802	0.9857	0.9789
0.0311	24.5482	72000	0.0926	0.9830	0.9918	0.9914	0.9961	0.9907	0.9853	0.9949	0.9839	0.9857	0.9793	0.9832
0.0564	25.9120	76000	0.0461	0.9900	0.9950	0.9950	0.9957	0.9937	0.9955	0.9952	0.9925	0.9907	0.9887	0.9883
0.0064	27.2758	80000	0.0458	0.9888	0.9945	0.9944	0.9958	0.9961	0.9908	0.9952	0.9869	0.9915	0.9853	0.9916
0.0023	28.6396	84000	0.0252	0.9929	0.9964	0.9964	0.9964	0.9977	0.9952	0.9962	0.9940	0.9938	0.9903	0.9936

Framework versions

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