segformer-b4-ade-finetuned-grCoastline

This model is a fine-tuned version of nvidia/segformer-b4-finetuned-ade-512-512 on the peldrak/grCoastline_512 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2116
• Mean Iou: 0.7207
• Mean Accuracy: 0.7864
• Overall Accuracy: 0.9441
• Accuracy Water: 0.9915
• Accuracy Whitewater: 0.0
• Accuracy Sediment: 0.9083
• Accuracy Other Natural Terrain: 0.8596
• Accuracy Vegetation: 0.8771
• Accuracy Development: 0.8701
• Accuracy Unknown: 0.9983
• Iou Water: 0.9580
• Iou Whitewater: 0.0
• Iou Sediment: 0.8691
• Iou Other Natural Terrain: 0.7108
• Iou Vegetation: 0.8236
• Iou Development: 0.6878
• Iou Unknown: 0.9957
• F1 Score: 0.9437

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Water	Accuracy Whitewater	Accuracy Sediment	Accuracy Other Natural Terrain	Accuracy Vegetation	Accuracy Development	Accuracy Unknown	Iou Water	Iou Whitewater	Iou Sediment	Iou Other Natural Terrain	Iou Vegetation	Iou Development	Iou Unknown	F1 Score
1.4532	0.24	20	1.3435	0.4211	0.5264	0.7441	0.7775	0.0	0.2537	0.5410	0.7502	0.3664	0.9961	0.6995	0.0	0.2230	0.3135	0.4349	0.3008	0.9758	0.7366
1.0467	0.49	40	0.8901	0.4799	0.5609	0.8306	0.9261	0.0	0.6581	0.3109	0.9513	0.0809	0.9992	0.8704	0.0	0.5553	0.3049	0.5627	0.0804	0.9860	0.8109
0.9494	0.73	60	0.6358	0.5209	0.5972	0.8680	0.9788	0.0	0.8070	0.4142	0.9622	0.0220	0.9964	0.9044	0.0	0.6693	0.4008	0.6578	0.0220	0.9921	0.8452
0.6118	0.98	80	0.5235	0.5454	0.6191	0.8834	0.9836	0.0	0.8558	0.5071	0.9635	0.0283	0.9951	0.9172	0.0	0.6969	0.4748	0.7090	0.0282	0.9914	0.8625
0.5458	1.22	100	0.4739	0.5846	0.6514	0.8875	0.9514	0.0	0.8431	0.5217	0.9639	0.2829	0.9967	0.9258	0.0	0.7593	0.4726	0.6684	0.2728	0.9930	0.8793
0.5935	1.46	120	0.3786	0.6090	0.6793	0.9055	0.9774	0.0	0.9195	0.7271	0.8854	0.2475	0.9980	0.9221	0.0	0.7820	0.5782	0.7474	0.2415	0.9921	0.8969
0.616	1.71	140	0.3630	0.6382	0.7043	0.9141	0.9845	0.0	0.9106	0.7205	0.8929	0.4243	0.9975	0.9132	0.0	0.7515	0.6445	0.7799	0.3859	0.9920	0.9090
0.6891	1.95	160	0.3563	0.6249	0.6897	0.9049	0.9912	0.0	0.8575	0.5156	0.9541	0.5139	0.9955	0.9010	0.0	0.7908	0.4895	0.7447	0.4577	0.9906	0.8978
0.3915	2.2	180	0.3039	0.6420	0.7094	0.9122	0.9690	0.0	0.8829	0.7779	0.8809	0.4559	0.9990	0.9296	0.0	0.8211	0.5859	0.7445	0.4225	0.9901	0.9093
0.478	2.44	200	0.2957	0.6661	0.7380	0.9201	0.9884	0.0	0.8179	0.8009	0.8854	0.6765	0.9973	0.9150	0.0	0.7697	0.6348	0.7986	0.5532	0.9916	0.9186
0.2891	2.68	220	0.2687	0.6781	0.7439	0.9258	0.9890	0.0	0.9161	0.7389	0.8977	0.6776	0.9881	0.9253	0.0	0.8384	0.6330	0.7898	0.5726	0.9874	0.9238
0.3432	2.93	240	0.2629	0.6725	0.7379	0.9246	0.9752	0.0	0.9444	0.7407	0.8825	0.6246	0.9976	0.9308	0.0	0.8389	0.6136	0.7804	0.5506	0.9935	0.9224
0.4379	3.17	260	0.2528	0.6787	0.7360	0.9283	0.9837	0.0	0.9170	0.7116	0.9280	0.6139	0.9974	0.9422	0.0	0.8382	0.6218	0.7870	0.5687	0.9931	0.9255
0.2129	3.41	280	0.2250	0.6775	0.7401	0.9264	0.9911	0.0	0.9001	0.6653	0.9301	0.7006	0.9936	0.9330	0.0	0.8456	0.6021	0.7849	0.5853	0.9918	0.9237
0.1941	3.66	300	0.2191	0.6943	0.7621	0.9336	0.9770	0.0	0.9426	0.8365	0.8619	0.7193	0.9974	0.9377	0.0	0.8296	0.6934	0.8076	0.5981	0.9938	0.9325
0.3357	3.9	320	0.2315	0.6841	0.7582	0.9267	0.9782	0.0	0.8671	0.7910	0.8801	0.7948	0.9961	0.9322	0.0	0.8089	0.6298	0.8011	0.6240	0.9926	0.9261
0.2456	4.15	340	0.2462	0.6904	0.7586	0.9303	0.9895	0.0	0.9229	0.6965	0.8897	0.8146	0.9970	0.9339	0.0	0.8563	0.6091	0.7915	0.6476	0.9947	0.9284
0.3712	4.39	360	0.2366	0.6918	0.7696	0.9307	0.9848	0.0	0.8491	0.8504	0.8689	0.8389	0.9951	0.9514	0.0	0.8085	0.6517	0.8102	0.6272	0.9935	0.9311
0.4034	4.63	380	0.2442	0.6854	0.7691	0.9254	0.9799	0.0	0.9505	0.8457	0.7848	0.8289	0.9940	0.9361	0.0	0.8287	0.6650	0.7533	0.6215	0.9930	0.9248
0.1794	4.88	400	0.2162	0.6951	0.7555	0.9338	0.9831	0.0	0.8868	0.7703	0.9249	0.7271	0.9967	0.9449	0.0	0.8392	0.6450	0.8101	0.6320	0.9945	0.9324
0.3036	5.12	420	0.2356	0.6869	0.7514	0.9303	0.9889	0.0	0.9184	0.6697	0.9145	0.7701	0.9983	0.9289	0.0	0.8442	0.6064	0.8075	0.6280	0.9932	0.9277
0.1948	5.37	440	0.2138	0.6963	0.7661	0.9333	0.9801	0.0	0.8889	0.8187	0.8823	0.7944	0.9982	0.9443	0.0	0.8363	0.6453	0.8165	0.6381	0.9939	0.9330
0.5839	5.61	460	0.2248	0.6998	0.7699	0.9350	0.9913	0.0	0.9310	0.7673	0.8695	0.8340	0.9966	0.9330	0.0	0.8511	0.6645	0.8143	0.6420	0.9936	0.9338
0.2039	5.85	480	0.2311	0.6949	0.7695	0.9300	0.9801	0.0	0.8510	0.8609	0.8619	0.8362	0.9967	0.9417	0.0	0.8106	0.6365	0.8076	0.6740	0.9942	0.9304
0.1818	6.1	500	0.2297	0.7037	0.7794	0.9338	0.9788	0.0	0.9420	0.8915	0.8103	0.8374	0.9955	0.9417	0.0	0.8651	0.6649	0.7871	0.6730	0.9942	0.9338
0.1637	6.34	520	0.1984	0.7105	0.7757	0.9389	0.9817	0.0	0.9503	0.8314	0.8634	0.8062	0.9969	0.9409	0.0	0.8601	0.6899	0.8123	0.6756	0.9946	0.9380
0.2034	6.59	540	0.1970	0.7170	0.7771	0.9426	0.9875	0.0	0.8953	0.8225	0.9205	0.8200	0.9939	0.9478	0.0	0.8602	0.7085	0.8317	0.6781	0.9926	0.9418
0.2673	6.83	560	0.2055	0.7111	0.7800	0.9400	0.9907	0.0	0.9323	0.8022	0.8739	0.8651	0.9957	0.9366	0.0	0.8675	0.6941	0.8262	0.6591	0.9939	0.9392
0.2446	7.07	580	0.2391	0.6769	0.7454	0.9286	0.9829	0.0	0.9409	0.9085	0.8298	0.5581	0.9979	0.9531	0.0	0.8470	0.6501	0.7880	0.5054	0.9944	0.9274
0.1747	7.32	600	0.2450	0.6993	0.7642	0.9351	0.9872	0.0	0.9200	0.7115	0.9121	0.8223	0.9967	0.9458	0.0	0.8690	0.6249	0.8060	0.6552	0.9939	0.9335
0.1262	7.56	620	0.2595	0.6955	0.7652	0.9315	0.9878	0.0	0.9453	0.7199	0.8657	0.8406	0.9970	0.9399	0.0	0.8588	0.6152	0.7856	0.6745	0.9946	0.9298
0.1223	7.8	640	0.2334	0.7028	0.7704	0.9339	0.9861	0.0	0.9214	0.8069	0.8572	0.8234	0.9977	0.9527	0.0	0.8672	0.6354	0.7841	0.6854	0.9948	0.9334
0.0915	8.05	660	0.2561	0.6879	0.7660	0.9258	0.9905	0.0	0.8363	0.8988	0.8221	0.8176	0.9967	0.9483	0.0	0.8053	0.6239	0.7819	0.6615	0.9946	0.9266
0.1095	8.29	680	0.2018	0.7179	0.7813	0.9431	0.9851	0.0	0.9306	0.7844	0.9086	0.8650	0.9950	0.9425	0.0	0.8742	0.6973	0.8350	0.6831	0.9933	0.9421
0.196	8.54	700	0.2125	0.7115	0.7834	0.9392	0.9800	0.0	0.9376	0.8437	0.8523	0.8722	0.9979	0.9515	0.0	0.8737	0.6855	0.8046	0.6707	0.9947	0.9389
0.1548	8.78	720	0.1893	0.7261	0.7833	0.9467	0.9830	0.0	0.9370	0.8452	0.9035	0.8170	0.9973	0.9545	0.0	0.8776	0.7329	0.8299	0.6936	0.9944	0.9458
0.1376	9.02	740	0.2467	0.7101	0.7706	0.9394	0.9868	0.0	0.9268	0.7615	0.9067	0.8150	0.9977	0.9536	0.0	0.8685	0.6624	0.8077	0.6831	0.9952	0.9381
0.1569	9.27	760	0.2038	0.7184	0.7823	0.9432	0.9904	0.0	0.9064	0.8209	0.8961	0.8651	0.9976	0.9505	0.0	0.8672	0.7048	0.8265	0.6845	0.9952	0.9425
0.1633	9.51	780	0.1953	0.7213	0.7820	0.9443	0.9882	0.0	0.9093	0.8436	0.8986	0.8371	0.9975	0.9499	0.0	0.8690	0.7129	0.8304	0.6919	0.9950	0.9436
0.1501	9.76	800	0.2605	0.7013	0.7688	0.9352	0.9799	0.0	0.8636	0.8899	0.8816	0.7689	0.9975	0.9573	0.0	0.8129	0.6623	0.8220	0.6595	0.9950	0.9353
0.2376	10.0	820	0.2071	0.7186	0.7850	0.9438	0.9922	0.0	0.9306	0.8279	0.8795	0.8683	0.9962	0.9505	0.0	0.8731	0.7145	0.8282	0.6695	0.9947	0.9431
0.0856	10.24	840	0.2190	0.7113	0.7689	0.9410	0.9878	0.0	0.9119	0.8175	0.9127	0.7559	0.9964	0.9583	0.0	0.8737	0.6801	0.8152	0.6568	0.9948	0.9401
0.2257	10.49	860	0.2133	0.7124	0.7847	0.9402	0.9871	0.0	0.9328	0.8304	0.8554	0.8895	0.9980	0.9565	0.0	0.8557	0.7005	0.8108	0.6686	0.9948	0.9398
0.2022	10.73	880	0.2517	0.7030	0.7746	0.9355	0.9845	0.0	0.8451	0.8379	0.8945	0.8631	0.9973	0.9514	0.0	0.8109	0.6656	0.8229	0.6754	0.9949	0.9353
0.0923	10.98	900	0.2118	0.7122	0.7717	0.9416	0.9863	0.0	0.8904	0.8660	0.9057	0.7555	0.9979	0.9563	0.0	0.8418	0.7016	0.8320	0.6582	0.9952	0.9409
0.1639	11.22	920	0.1961	0.7244	0.7817	0.9471	0.9835	0.0	0.9354	0.8432	0.9118	0.8013	0.9967	0.9584	0.0	0.8649	0.7337	0.8397	0.6790	0.9951	0.9462
0.0676	11.46	940	0.2238	0.7135	0.7829	0.9421	0.9885	0.0	0.9482	0.7915	0.8765	0.8784	0.9973	0.9503	0.0	0.8653	0.7000	0.8282	0.6555	0.9952	0.9413
0.056	11.71	960	0.1908	0.7284	0.7848	0.9478	0.9862	0.0	0.9206	0.8434	0.9143	0.8319	0.9973	0.9583	0.0	0.8724	0.7308	0.8375	0.7048	0.9948	0.9470
0.1205	11.95	980	0.2000	0.7248	0.7914	0.9457	0.9844	0.0	0.9377	0.8256	0.8807	0.9132	0.9986	0.9581	0.0	0.8681	0.7249	0.8265	0.7016	0.9947	0.9450
0.0429	12.2	1000	0.1937	0.7285	0.7865	0.9476	0.9856	0.0	0.9298	0.8327	0.9064	0.8529	0.9982	0.9590	0.0	0.8687	0.7326	0.8335	0.7108	0.9952	0.9468
0.0992	12.44	1020	0.2081	0.7225	0.7858	0.9448	0.9863	0.0	0.9394	0.8478	0.8780	0.8522	0.9972	0.9576	0.0	0.8668	0.7191	0.8230	0.6955	0.9953	0.9441
0.089	12.68	1040	0.2044	0.7243	0.7812	0.9466	0.9832	0.0	0.9240	0.8403	0.9128	0.8095	0.9987	0.9561	0.0	0.8724	0.7274	0.8351	0.6844	0.9949	0.9457
0.1252	12.93	1060	0.2146	0.7138	0.7750	0.9417	0.9824	0.0	0.9125	0.8768	0.8895	0.7662	0.9973	0.9573	0.0	0.8521	0.6993	0.8267	0.6659	0.9950	0.9412
0.3201	13.17	1080	0.2169	0.7197	0.7849	0.9439	0.9796	0.0	0.9489	0.8613	0.8713	0.8352	0.9977	0.9580	0.0	0.8636	0.7225	0.8209	0.6774	0.9955	0.9434
0.101	13.41	1100	0.2023	0.7268	0.7848	0.9482	0.9871	0.0	0.9314	0.8350	0.9111	0.8315	0.9978	0.9587	0.0	0.8702	0.7416	0.8412	0.6806	0.9952	0.9473
0.0522	13.66	1120	0.2116	0.7207	0.7864	0.9441	0.9915	0.0	0.9083	0.8596	0.8771	0.8701	0.9983	0.9580	0.0	0.8691	0.7108	0.8236	0.6878	0.9957	0.9437

Framework versions

• Transformers 4.38.1
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2