segformer-b5-finetuned-ade20k-hgo-coord_40epochs_distortion_ver2_global_norm_with_void_4

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

• Loss: 0.6653
• Mean Iou: 0.7725
• Mean Accuracy: 0.8702
• Overall Accuracy: 0.8824
• Accuracy [0,0]: 0.8550
• Accuracy [0,1]: 0.8883
• Accuracy [1,0]: 0.9019
• Accuracy [1,1]: 0.8817
• Accuracy [0,2]: 0.8976
• Accuracy [0,3]: 0.9033
• Accuracy [1,2]: 0.8715
• Accuracy [1,3]: 0.9091
• Accuracy [2,0]: 0.8286
• Accuracy [2,1]: 0.8755
• Accuracy [2,2]: 0.8668
• Accuracy [2,3]: 0.8119
• Accuracy [3,0]: 0.8624
• Accuracy [3,1]: 0.7922
• Accuracy [3,2]: 0.8500
• Accuracy [3,3]: 0.8287
• Accuracy Void: 0.9695
• Iou [0,0]: 0.7906
• Iou [0,1]: 0.8047
• Iou [1,0]: 0.7816
• Iou [1,1]: 0.8141
• Iou [0,2]: 0.8098
• Iou [0,3]: 0.7654
• Iou [1,2]: 0.7771
• Iou [1,3]: 0.7698
• Iou [2,0]: 0.7262
• Iou [2,1]: 0.7632
• Iou [2,2]: 0.7299
• Iou [2,3]: 0.7208
• Iou [3,0]: 0.7854
• Iou [3,1]: 0.7184
• Iou [3,2]: 0.7428
• Iou [3,3]: 0.7067
• Iou Void: 0.9263

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: 160

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy [0,0]	Accuracy [0,1]	Accuracy [1,0]	Accuracy [1,1]	Accuracy [0,2]	Accuracy [0,3]	Accuracy [1,2]	Accuracy [1,3]	Accuracy [2,0]	Accuracy [2,1]	Accuracy [2,2]	Accuracy [2,3]	Accuracy [3,0]	Accuracy [3,1]	Accuracy [3,2]	Accuracy [3,3]	Accuracy Void	Iou [0,0]	Iou [0,1]	Iou [1,0]	Iou [1,1]	Iou [0,2]	Iou [0,3]	Iou [1,2]	Iou [1,3]	Iou [2,0]	Iou [2,1]	Iou [2,2]	Iou [2,3]	Iou [3,0]	Iou [3,1]	Iou [3,2]	Iou [3,3]	Iou Void
1.1886	7.3260	4000	1.2673	0.3996	0.5640	0.6016	0.4741	0.5454	0.5928	0.6399	0.5156	0.4641	0.4718	0.5480	0.4585	0.5101	0.5382	0.5261	0.6719	0.5590	0.4871	0.6623	0.9223	0.4093	0.4143	0.3984	0.4226	0.4037	0.3569	0.3321	0.4267	0.3569	0.3466	0.3401	0.3917	0.4129	0.3384	0.3123	0.2921	0.8383
1.2869	14.6520	8000	0.9750	0.5116	0.6709	0.7020	0.6675	0.6769	0.7363	0.7205	0.6219	0.6666	0.4876	0.7955	0.6715	0.6206	0.5169	0.7125	0.7610	0.4896	0.6361	0.7068	0.9174	0.5488	0.5559	0.5446	0.4579	0.5345	0.5201	0.4101	0.5248	0.4857	0.4404	0.4303	0.4828	0.5633	0.4133	0.5004	0.4308	0.8539
1.2291	21.9780	12000	0.8046	0.5963	0.7469	0.7658	0.7366	0.7485	0.7818	0.7346	0.7464	0.8049	0.6572	0.8151	0.6797	0.7084	0.7394	0.7938	0.7353	0.6297	0.7787	0.7009	0.9068	0.6130	0.6405	0.6005	0.6198	0.6201	0.5553	0.5344	0.6296	0.5199	0.5460	0.4700	0.6153	0.6217	0.5344	0.5849	0.5573	0.8750
0.3104	29.3040	16000	0.6399	0.6582	0.7930	0.8091	0.7850	0.8352	0.8091	0.8373	0.8137	0.8251	0.7227	0.7906	0.8229	0.7740	0.7435	0.8859	0.8129	0.6392	0.7057	0.7630	0.9157	0.6952	0.6891	0.6964	0.6733	0.6595	0.6223	0.6479	0.6870	0.6201	0.6079	0.6430	0.6449	0.6967	0.5473	0.5996	0.5790	0.8810
0.3486	36.6300	20000	0.6188	0.6709	0.8015	0.8193	0.7634	0.8269	0.8616	0.8582	0.8198	0.8505	0.6733	0.8336	0.8299	0.8104	0.7407	0.8566	0.7730	0.6485	0.7307	0.8084	0.9401	0.7015	0.7169	0.6902	0.7149	0.6658	0.6510	0.6201	0.6931	0.6274	0.6539	0.6000	0.6483	0.6773	0.5972	0.6464	0.6074	0.8940
0.2437	43.9560	24000	0.6233	0.6775	0.8026	0.8251	0.8422	0.8838	0.8424	0.8420	0.8351	0.8449	0.6714	0.8251	0.8146	0.8303	0.7672	0.8111	0.7900	0.6462	0.7008	0.7325	0.9644	0.7360	0.6873	0.6956	0.7231	0.6931	0.6536	0.6321	0.7337	0.6171	0.6415	0.5926	0.6804	0.7243	0.5894	0.5893	0.6230	0.9052
0.1864	51.2821	28000	0.5680	0.7150	0.8333	0.8473	0.8205	0.8365	0.8748	0.8464	0.8739	0.8255	0.8341	0.8842	0.7700	0.7983	0.8625	0.7927	0.8406	0.7202	0.8199	0.8125	0.9538	0.7419	0.7433	0.7285	0.7755	0.6894	0.6936	0.7199	0.7695	0.6154	0.6969	0.6473	0.6570	0.7580	0.6683	0.6683	0.6762	0.9059
0.1692	58.6081	32000	0.5921	0.7288	0.8426	0.8558	0.8258	0.8749	0.8927	0.8481	0.8773	0.8747	0.8187	0.8771	0.7955	0.8649	0.7956	0.7949	0.8335	0.7759	0.8405	0.7863	0.9475	0.7648	0.7547	0.7451	0.7684	0.7623	0.7160	0.7242	0.7323	0.6573	0.6880	0.6486	0.7025	0.7500	0.6799	0.7119	0.6784	0.9057
0.4861	65.9341	36000	0.5194	0.7383	0.8482	0.8616	0.8336	0.8530	0.8778	0.8545	0.8688	0.8927	0.8369	0.8942	0.8213	0.8737	0.8223	0.8568	0.8525	0.6965	0.8116	0.8126	0.9609	0.7682	0.7622	0.7594	0.7796	0.7457	0.6981	0.7502	0.7548	0.6797	0.7048	0.6785	0.7433	0.7979	0.6522	0.7041	0.6543	0.9186
0.0915	73.2601	40000	0.5566	0.7394	0.8480	0.8621	0.8206	0.8965	0.9048	0.8691	0.8445	0.8811	0.8250	0.9031	0.8086	0.8207	0.8112	0.8027	0.8587	0.7725	0.8267	0.8175	0.9533	0.7485	0.7813	0.7311	0.7713	0.7492	0.7206	0.7520	0.7441	0.6908	0.7191	0.7050	0.7131	0.7688	0.6863	0.6970	0.6761	0.9154
0.077	80.5861	44000	0.5688	0.7463	0.8535	0.8664	0.8592	0.8755	0.9036	0.8583	0.8760	0.8869	0.8099	0.9010	0.8338	0.8629	0.7998	0.8509	0.8282	0.7651	0.8461	0.8025	0.9504	0.7777	0.7797	0.7702	0.7893	0.7733	0.7193	0.7441	0.7597	0.6706	0.7043	0.6729	0.7524	0.7556	0.7023	0.7405	0.6601	0.9144
0.157	87.9121	48000	0.5899	0.7461	0.8530	0.8667	0.8567	0.8936	0.9126	0.8858	0.8789	0.8671	0.8358	0.8843	0.7829	0.8759	0.8621	0.7755	0.8669	0.7841	0.7996	0.7827	0.9564	0.7788	0.7744	0.7384	0.7894	0.7758	0.7410	0.7388	0.7349	0.6856	0.7261	0.7241	0.7141	0.7745	0.6944	0.7012	0.6713	0.9202
0.1121	95.2381	52000	0.5786	0.7497	0.8572	0.8687	0.7989	0.8786	0.9104	0.8817	0.8724	0.8860	0.8292	0.8782	0.8114	0.8692	0.8686	0.8451	0.8437	0.8010	0.8048	0.8363	0.9572	0.7612	0.7862	0.7810	0.7833	0.7666	0.7264	0.7541	0.7763	0.7090	0.7208	0.6813	0.7268	0.7698	0.6926	0.6978	0.6915	0.9200
0.1639	102.5641	56000	0.6080	0.7492	0.8558	0.8690	0.8640	0.8562	0.8978	0.8556	0.8780	0.8913	0.8356	0.8889	0.8292	0.8292	0.8665	0.8356	0.8422	0.7435	0.8396	0.8296	0.9651	0.8072	0.7678	0.7487	0.7867	0.7676	0.7490	0.7302	0.7669	0.6984	0.7091	0.6520	0.7207	0.7901	0.6841	0.7341	0.7021	0.9220
0.1274	109.8901	60000	0.5982	0.7551	0.8589	0.8722	0.8467	0.8706	0.9042	0.8514	0.8906	0.9028	0.8519	0.9049	0.7823	0.8592	0.8388	0.8417	0.8580	0.7620	0.8412	0.8282	0.9673	0.7838	0.7805	0.7717	0.8013	0.7760	0.7264	0.7607	0.7828	0.6788	0.7438	0.6709	0.7439	0.7875	0.7008	0.7342	0.6706	0.9241
0.0471	117.2161	64000	0.6311	0.7516	0.8551	0.8701	0.8204	0.8726	0.9208	0.8911	0.8795	0.8946	0.8237	0.9084	0.8002	0.8610	0.8294	0.8125	0.8272	0.7370	0.8589	0.8228	0.9765	0.7672	0.7862	0.7681	0.7937	0.7888	0.7340	0.7439	0.7461	0.6789	0.7402	0.7042	0.7235	0.7818	0.6896	0.7305	0.6793	0.9216
0.1196	124.5421	68000	0.6434	0.7574	0.8591	0.8729	0.8381	0.8515	0.9121	0.8759	0.8960	0.9228	0.8405	0.9020	0.8199	0.8498	0.8307	0.8084	0.8578	0.7502	0.8544	0.8232	0.9713	0.7775	0.7835	0.7660	0.7959	0.7943	0.7476	0.7604	0.7338	0.7099	0.7565	0.7205	0.7249	0.7782	0.6936	0.7448	0.6688	0.9201
0.0608	131.8681	72000	0.6561	0.7643	0.8649	0.8778	0.8422	0.8783	0.9092	0.8826	0.9052	0.8930	0.8790	0.8970	0.7703	0.8836	0.8538	0.8109	0.8547	0.7824	0.8676	0.8203	0.9735	0.7761	0.8057	0.7746	0.8099	0.7989	0.7626	0.7788	0.7617	0.6689	0.7437	0.7134	0.7253	0.7774	0.7210	0.7450	0.7057	0.9240
0.3017	139.1941	76000	0.6601	0.7695	0.8682	0.8806	0.8440	0.8853	0.9105	0.8975	0.8843	0.8977	0.8319	0.9032	0.8348	0.8918	0.8607	0.7982	0.8622	0.7801	0.8673	0.8393	0.9705	0.7834	0.7974	0.7818	0.8137	0.7982	0.7578	0.7558	0.7736	0.7101	0.7564	0.7204	0.7175	0.7931	0.7169	0.7660	0.7131	0.9255
0.0676	146.5201	80000	0.6446	0.7707	0.8687	0.8815	0.8372	0.8897	0.9133	0.8882	0.9012	0.8940	0.8804	0.9149	0.8135	0.8804	0.8601	0.7968	0.8520	0.7825	0.8531	0.8347	0.9754	0.7825	0.8057	0.7876	0.8081	0.8047	0.7625	0.7868	0.7803	0.7166	0.7450	0.7153	0.7174	0.7854	0.7197	0.7427	0.7183	0.9234
0.0312	153.8462	84000	0.6653	0.7725	0.8702	0.8824	0.8550	0.8883	0.9019	0.8817	0.8976	0.9033	0.8715	0.9091	0.8286	0.8755	0.8668	0.8119	0.8624	0.7922	0.8500	0.8287	0.9695	0.7906	0.8047	0.7816	0.8141	0.8098	0.7654	0.7771	0.7698	0.7262	0.7632	0.7299	0.7208	0.7854	0.7184	0.7428	0.7067	0.9263

Framework versions

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