segformer-b0-scene-parse-150

This model is a fine-tuned version of nvidia/mit-b0 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.9881
• Mean Iou: 0.0898
• Mean Accuracy: 0.1508
• Overall Accuracy: 0.6092

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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy
4.8641	1.0	20	4.9027	0.0114	0.0472	0.2362
4.4897	2.0	40	4.5156	0.0223	0.0976	0.3997
4.1506	3.0	60	3.9759	0.0415	0.1049	0.4639
3.9239	4.0	80	3.5152	0.0513	0.1151	0.4998
3.7073	5.0	100	3.1904	0.0662	0.1180	0.5333
3.4644	6.0	120	3.0165	0.0694	0.1181	0.5548
3.3538	7.0	140	2.9451	0.0698	0.1189	0.5437
2.9297	8.0	160	2.8302	0.0794	0.1252	0.5704
3.0257	9.0	180	2.6410	0.0801	0.1307	0.5749
2.8081	10.0	200	2.4851	0.0803	0.1259	0.5883
2.6384	11.0	220	2.4904	0.0853	0.1318	0.5817
2.5555	12.0	240	2.3738	0.0874	0.1296	0.6095
2.3822	13.0	260	2.3239	0.0864	0.1317	0.5924
2.4267	14.0	280	2.2778	0.0843	0.1286	0.6019
2.3553	15.0	300	2.2917	0.0836	0.1301	0.5899
2.1810	16.0	320	2.2187	0.0873	0.1388	0.5972
2.2971	17.0	340	2.1589	0.0890	0.1423	0.6107
2.2366	18.0	360	2.2463	0.0866	0.1484	0.5969
2.2299	19.0	380	2.0700	0.0900	0.1415	0.6184
1.9433	20.0	400	2.0822	0.0891	0.1457	0.6177
1.9851	21.0	420	2.1299	0.0885	0.1484	0.5968
1.9428	22.0	440	2.0912	0.0856	0.1416	0.6082
1.6539	23.0	460	2.2378	0.0866	0.1464	0.5902
1.9056	24.0	480	2.0942	0.0896	0.1423	0.6142
1.6493	25.0	500	2.0635	0.0915	0.1477	0.6160
1.7495	26.0	520	2.0396	0.0920	0.1483	0.6179
1.5518	27.0	540	2.0270	0.0904	0.1485	0.6098
1.6620	28.0	560	2.0564	0.0931	0.1585	0.6098
1.5614	29.0	580	2.0542	0.0942	0.1536	0.6106
1.5037	30.0	600	2.0698	0.0874	0.1477	0.5995
1.3493	31.0	620	2.0744	0.0880	0.1539	0.5926
1.6179	32.0	640	2.0404	0.0906	0.1493	0.6081
1.5231	33.0	660	2.0153	0.0907	0.1506	0.6085
1.4016	34.0	680	2.1269	0.0860	0.1513	0.5846
1.5147	35.0	700	2.0160	0.0920	0.1472	0.6120
1.4723	36.0	720	1.9838	0.0948	0.1525	0.6211
1.3963	37.0	740	2.0161	0.0886	0.1514	0.6108
1.5678	38.0	760	2.0203	0.0887	0.1526	0.6133
1.4418	39.0	780	1.9610	0.0932	0.1470	0.6112
1.4737	40.0	800	1.9609	0.0912	0.1466	0.6106
1.3634	41.0	820	1.9471	0.0942	0.1509	0.6175
1.4730	42.0	840	1.9611	0.0939	0.1520	0.6152
1.4209	43.0	860	1.9680	0.0901	0.1465	0.6044
1.2518	44.0	880	1.9577	0.0918	0.1502	0.6121
1.2380	45.0	900	1.9631	0.0909	0.1516	0.6097
1.3753	46.0	920	1.9620	0.0919	0.1489	0.6090
1.2756	47.0	940	1.9659	0.0909	0.1499	0.6104
1.3696	48.0	960	1.9568	0.0898	0.1495	0.6098
1.2820	49.0	980	2.0069	0.0911	0.1524	0.6072
1.3792	50.0	1000	1.9881	0.0898	0.1508	0.6092

Framework versions

• Transformers 5.8.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2