JCAI2000/segformer-b5-finetuned-100by100PNG-50epochs-attempt2-100epochs-backgroundclass-2

segformer-b3-finetuned-100by100PNG-50epochs-attempt2-100epochs-backgroundclass

This model is a fine-tuned version of nvidia/mit-b5 on the JCAI2000/100By100BranchPNG dataset. It achieves the following results on the evaluation set:

• Loss: 0.1497
• Mean Iou: 0.8933
• Mean Accuracy: 0.9531
• Overall Accuracy: 0.9662
• Accuracy Background: 0.9732
• Accuracy Branch: 0.9330
• Iou Background: 0.9597
• Iou Branch: 0.8270

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Branch	Iou Background	Iou Branch
0.2055	1.05	20	0.2925	0.8151	0.9469	0.9320	0.9242	0.9695	0.9183	0.7118
0.1549	2.11	40	0.1328	0.8802	0.9311	0.9628	0.9796	0.8825	0.9561	0.8043
0.0735	3.16	60	0.1178	0.8804	0.9512	0.9613	0.9666	0.9357	0.9538	0.8070
0.0636	4.21	80	0.0844	0.8966	0.9368	0.9686	0.9854	0.8881	0.9629	0.8303
0.0546	5.26	100	0.1099	0.8969	0.9526	0.9676	0.9756	0.9297	0.9614	0.8325
0.0567	6.32	120	0.1012	0.8996	0.9500	0.9688	0.9788	0.9213	0.9629	0.8364
0.0515	7.37	140	0.1137	0.8935	0.9462	0.9668	0.9777	0.9147	0.9605	0.8265
0.052	8.42	160	0.0987	0.8914	0.9317	0.9670	0.9858	0.8776	0.9611	0.8217
0.0358	9.47	180	0.1167	0.8978	0.9581	0.9676	0.9726	0.9435	0.9613	0.8344
0.0254	10.53	200	0.0767	0.9111	0.9519	0.9729	0.9840	0.9197	0.9678	0.8545
0.0483	11.58	220	0.0953	0.9037	0.9524	0.9701	0.9795	0.9253	0.9645	0.8429
0.0285	12.63	240	0.0904	0.9026	0.9490	0.9700	0.9811	0.9169	0.9643	0.8409
0.0389	13.68	260	0.0902	0.9025	0.9472	0.9701	0.9821	0.9123	0.9644	0.8406
0.0473	14.74	280	0.0852	0.9084	0.9522	0.9719	0.9823	0.9220	0.9665	0.8502
0.0266	15.79	300	0.0983	0.8985	0.9409	0.9690	0.9839	0.8979	0.9633	0.8337
0.0233	16.84	320	0.0965	0.9052	0.9601	0.9702	0.9756	0.9447	0.9644	0.8460
0.0257	17.89	340	0.0941	0.9039	0.9550	0.9701	0.9781	0.9319	0.9643	0.8434
0.0352	18.95	360	0.0855	0.9043	0.9483	0.9706	0.9824	0.9142	0.9651	0.8435
0.1941	20.0	380	0.0946	0.9045	0.9509	0.9706	0.9809	0.9210	0.9650	0.8441
0.0325	21.05	400	0.0972	0.8973	0.9449	0.9683	0.9807	0.9092	0.9624	0.8323
0.0159	22.11	420	0.0828	0.9081	0.9528	0.9717	0.9817	0.9239	0.9664	0.8498
0.0175	23.16	440	0.1061	0.8995	0.9491	0.9688	0.9793	0.9188	0.9629	0.8360
0.0281	24.21	460	0.1090	0.8969	0.9516	0.9677	0.9761	0.9271	0.9615	0.8323
0.0177	25.26	480	0.1122	0.8983	0.9547	0.9680	0.9750	0.9343	0.9618	0.8347
0.0228	26.32	500	0.1088	0.8957	0.9546	0.9670	0.9736	0.9357	0.9606	0.8307
0.0348	27.37	520	0.0933	0.9059	0.9524	0.9710	0.9808	0.9241	0.9654	0.8464
0.0177	28.42	540	0.1053	0.9025	0.9527	0.9697	0.9787	0.9268	0.9639	0.8411
0.0182	29.47	560	0.1039	0.8992	0.9473	0.9688	0.9802	0.9143	0.9630	0.8355
0.0171	30.53	580	0.1117	0.8991	0.9555	0.9682	0.9750	0.9360	0.9621	0.8361
0.0275	31.58	600	0.1142	0.8935	0.9497	0.9665	0.9754	0.9241	0.9601	0.8268
0.0186	32.63	620	0.1065	0.9024	0.9524	0.9697	0.9788	0.9261	0.9639	0.8408
0.0173	33.68	640	0.1081	0.8986	0.9529	0.9682	0.9764	0.9294	0.9621	0.8351
0.015	34.74	660	0.1243	0.8935	0.9530	0.9663	0.9733	0.9327	0.9598	0.8272
0.0183	35.79	680	0.1120	0.9005	0.9500	0.9691	0.9792	0.9209	0.9633	0.8377
0.0248	36.84	700	0.1185	0.8962	0.9517	0.9674	0.9757	0.9277	0.9611	0.8312
0.0104	37.89	720	0.1136	0.8975	0.9506	0.9680	0.9771	0.9241	0.9619	0.8332
0.0481	38.95	740	0.1127	0.9010	0.9528	0.9691	0.9778	0.9277	0.9632	0.8388
0.0153	40.0	760	0.1101	0.9019	0.9537	0.9694	0.9777	0.9297	0.9635	0.8402
0.0143	41.05	780	0.1105	0.9032	0.9558	0.9698	0.9771	0.9345	0.9639	0.8425
0.0104	42.11	800	0.1122	0.8986	0.9428	0.9689	0.9827	0.9028	0.9631	0.8340
0.0172	43.16	820	0.1097	0.9041	0.9540	0.9702	0.9788	0.9291	0.9645	0.8437
0.0371	44.21	840	0.1064	0.9011	0.9503	0.9693	0.9794	0.9212	0.9635	0.8387
0.0221	45.26	860	0.1150	0.9004	0.9515	0.9690	0.9783	0.9247	0.9631	0.8377
0.0186	46.32	880	0.1228	0.8958	0.9518	0.9672	0.9754	0.9282	0.9610	0.8306
0.0119	47.37	900	0.1205	0.8980	0.9525	0.9680	0.9762	0.9288	0.9619	0.8340
0.0113	48.42	920	0.1133	0.8998	0.9502	0.9688	0.9787	0.9216	0.9629	0.8366
0.0121	49.47	940	0.1145	0.8993	0.9490	0.9688	0.9792	0.9188	0.9629	0.8358
0.0263	50.53	960	0.1168	0.8977	0.9542	0.9678	0.9750	0.9334	0.9616	0.8338
0.0093	51.58	980	0.1213	0.8940	0.9534	0.9664	0.9733	0.9334	0.9600	0.8280
0.0193	52.63	1000	0.1241	0.8971	0.9507	0.9678	0.9769	0.9246	0.9617	0.8326
0.0139	53.68	1020	0.1263	0.8962	0.9546	0.9672	0.9739	0.9353	0.9609	0.8316
0.012	54.74	1040	0.1252	0.8952	0.9504	0.9671	0.9760	0.9247	0.9609	0.8296
0.008	55.79	1060	0.1219	0.8986	0.9516	0.9683	0.9772	0.9260	0.9623	0.8349
0.0092	56.84	1080	0.1290	0.8995	0.9552	0.9684	0.9754	0.9349	0.9623	0.8366
0.015	57.89	1100	0.1243	0.8989	0.9545	0.9682	0.9755	0.9335	0.9621	0.8358
0.0126	58.95	1120	0.1214	0.8977	0.9541	0.9678	0.9751	0.9331	0.9616	0.8337
0.0212	60.0	1140	0.1298	0.8953	0.9542	0.9669	0.9736	0.9347	0.9605	0.8301
0.0192	61.05	1160	0.1341	0.8930	0.9518	0.9661	0.9737	0.9299	0.9597	0.8262
0.0136	62.11	1180	0.1327	0.8970	0.9528	0.9676	0.9754	0.9302	0.9614	0.8325
0.0131	63.16	1200	0.1233	0.8997	0.9549	0.9685	0.9757	0.9340	0.9624	0.8369
0.0135	64.21	1220	0.1301	0.8957	0.9542	0.9670	0.9738	0.9345	0.9607	0.8307
0.0228	65.26	1240	0.1274	0.8979	0.9524	0.9680	0.9762	0.9285	0.9618	0.8339
0.0138	66.32	1260	0.1336	0.8965	0.9520	0.9675	0.9757	0.9283	0.9613	0.8318
0.0127	67.37	1280	0.1278	0.8980	0.9519	0.9681	0.9767	0.9271	0.9620	0.8341
0.0107	68.42	1300	0.1293	0.8970	0.9530	0.9676	0.9753	0.9308	0.9614	0.8327
0.0278	69.47	1320	0.1413	0.8926	0.9534	0.9659	0.9725	0.9343	0.9593	0.8258
0.0159	70.53	1340	0.1360	0.8953	0.9522	0.9670	0.9748	0.9296	0.9607	0.8298
0.0105	71.58	1360	0.1319	0.8972	0.9537	0.9676	0.9750	0.9324	0.9614	0.8330
0.0168	72.63	1380	0.1343	0.8942	0.9533	0.9665	0.9735	0.9331	0.9601	0.8283
0.0156	73.68	1400	0.1357	0.8950	0.9516	0.9669	0.9751	0.9281	0.9606	0.8294
0.0109	74.74	1420	0.1446	0.8905	0.9524	0.9652	0.9719	0.9328	0.9585	0.8226
0.0168	75.79	1440	0.1339	0.8958	0.9533	0.9671	0.9745	0.9320	0.9608	0.8308
0.0252	76.84	1460	0.1355	0.8935	0.9532	0.9662	0.9731	0.9333	0.9597	0.8272
0.0109	77.89	1480	0.1388	0.8932	0.9533	0.9661	0.9729	0.9338	0.9596	0.8267
0.0109	78.95	1500	0.1404	0.8924	0.9519	0.9659	0.9734	0.9305	0.9594	0.8255
0.0112	80.0	1520	0.1424	0.8921	0.9535	0.9657	0.9722	0.9349	0.9591	0.8251
0.0094	81.05	1540	0.1451	0.8924	0.9524	0.9659	0.9730	0.9317	0.9593	0.8254
0.007	82.11	1560	0.1457	0.8931	0.9527	0.9661	0.9732	0.9322	0.9596	0.8266
0.0119	83.16	1580	0.1424	0.8927	0.9520	0.9661	0.9735	0.9306	0.9595	0.8259
0.0153	84.21	1600	0.1535	0.8909	0.9530	0.9653	0.9718	0.9342	0.9586	0.8233
0.0104	85.26	1620	0.1452	0.8921	0.9529	0.9658	0.9725	0.9333	0.9592	0.8251
0.0101	86.32	1640	0.1503	0.8910	0.9536	0.9653	0.9714	0.9358	0.9586	0.8235
0.009	87.37	1660	0.1508	0.8925	0.9532	0.9659	0.9725	0.9339	0.9593	0.8257
0.0073	88.42	1680	0.1419	0.8949	0.9528	0.9668	0.9742	0.9315	0.9604	0.8293
0.0137	89.47	1700	0.1437	0.8942	0.9526	0.9666	0.9739	0.9313	0.9601	0.8282
0.0061	90.53	1720	0.1474	0.8928	0.9523	0.9660	0.9733	0.9313	0.9595	0.8260
0.0132	91.58	1740	0.1408	0.8935	0.9522	0.9663	0.9738	0.9306	0.9598	0.8271
0.0089	92.63	1760	0.1468	0.8933	0.9527	0.9662	0.9734	0.9320	0.9597	0.8268
0.0141	93.68	1780	0.1458	0.8930	0.9529	0.9661	0.9731	0.9328	0.9596	0.8265
0.0132	94.74	1800	0.1442	0.8934	0.9528	0.9662	0.9733	0.9324	0.9597	0.8270
0.0109	95.79	1820	0.1445	0.8928	0.9529	0.9660	0.9730	0.9327	0.9595	0.8262
0.0248	96.84	1840	0.1391	0.8938	0.9529	0.9664	0.9736	0.9322	0.9599	0.8276
0.0074	97.89	1860	0.1424	0.8938	0.9531	0.9664	0.9735	0.9327	0.9599	0.8277
0.0072	98.95	1880	0.1465	0.8931	0.9534	0.9661	0.9728	0.9339	0.9596	0.8266
0.0183	100.0	1900	0.1497	0.8933	0.9531	0.9662	0.9732	0.9330	0.9597	0.8270

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.4
• Tokenizers 0.13.3