segformer-b2-seed63-apr-13-v1

This model is a fine-tuned version of nvidia/mit-b3 on the unreal-hug/REAL_DATASET_SEG_401_6_lbls dataset. It achieves the following results on the evaluation set:

• Loss: 1.7138
• Mean Iou: 0.1266
• Mean Accuracy: 0.2136
• Overall Accuracy: 0.4273
• Accuracy Unlabeled: nan
• Accuracy Lv: 0.6939
• Accuracy Rv: 0.0982
• Accuracy Ra: 0.1706
• Accuracy La: 0.5041
• Accuracy Vs: 0.0
• Accuracy As: 0.0
• Accuracy Mk: 0.0
• Accuracy Tk: nan
• Accuracy Asd: 0.0557
• Accuracy Vsd: 0.2283
• Accuracy Ak: 0.3849
• Iou Unlabeled: 0.0
• Iou Lv: 0.4965
• Iou Rv: 0.0899
• Iou Ra: 0.1288
• Iou La: 0.2845
• Iou Vs: 0.0
• Iou As: 0.0
• Iou Mk: 0.0
• Iou Tk: 0.0
• Iou Asd: 0.0462
• Iou Vsd: 0.1513
• Iou Ak: 0.3225

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: 1e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Lv	Accuracy Rv	Accuracy Ra	Accuracy La	Accuracy Vs	Accuracy As	Accuracy Mk	Accuracy Tk	Accuracy Asd	Accuracy Vsd	Accuracy Ak	Iou Unlabeled	Iou Lv	Iou Rv	Iou Ra	Iou La	Iou Vs	Iou As	Iou Mk	Iou Tk	Iou Asd	Iou Vsd	Iou Ak
2.5423	2.5	100	2.6367	0.0332	0.0976	0.0951	nan	0.0612	0.0642	0.0301	0.1898	0.0	0.0	0.0086	nan	0.0495	0.4697	0.1033	0.0	0.0573	0.0485	0.0262	0.1021	0.0	0.0	0.0019	0.0	0.0204	0.0612	0.0812
2.3042	5.0	200	2.3925	0.0604	0.1412	0.1975	nan	0.2435	0.0655	0.1292	0.2869	0.0	0.0	0.0046	nan	0.0669	0.4894	0.1258	0.0	0.2144	0.0516	0.1074	0.1515	0.0	0.0	0.0017	0.0	0.0243	0.0670	0.1063
2.0869	7.5	300	2.2183	0.0932	0.1839	0.3354	nan	0.5208	0.0717	0.1836	0.4192	0.0	0.0	0.0006	nan	0.0768	0.3608	0.2060	0.0	0.4077	0.0617	0.1436	0.2158	0.0	0.0	0.0003	0.0	0.0358	0.0787	0.1746
2.0559	10.0	400	2.0298	0.1110	0.2055	0.3886	nan	0.6144	0.1027	0.1815	0.4598	0.0	0.0	0.0005	nan	0.0909	0.3011	0.3041	0.0	0.4559	0.0880	0.1400	0.2409	0.0	0.0	0.0003	0.0	0.0534	0.1001	0.2538
1.9554	12.5	500	1.8871	0.1189	0.2111	0.4100	nan	0.6561	0.1004	0.1647	0.4900	0.0	0.0	0.0009	nan	0.0763	0.2611	0.3619	0.0	0.4739	0.0896	0.1263	0.2616	0.0	0.0	0.0007	0.0	0.0531	0.1207	0.3015
2.0181	15.0	600	1.7720	0.1247	0.2139	0.4199	nan	0.6735	0.1008	0.1723	0.4898	0.0	0.0	0.0	nan	0.0706	0.2349	0.3972	0.0	0.4860	0.0912	0.1293	0.2720	0.0	0.0	0.0	0.0	0.0532	0.1386	0.3256
1.6723	17.5	700	1.7386	0.1258	0.2129	0.4251	nan	0.6860	0.1011	0.1724	0.5062	0.0	0.0	0.0	nan	0.0615	0.2167	0.3848	0.0	0.4927	0.0917	0.1304	0.2814	0.0	0.0	0.0	0.0	0.0488	0.1426	0.3221
1.5613	20.0	800	1.7751	0.1269	0.2151	0.4322	nan	0.7050	0.1020	0.1730	0.5066	0.0	0.0	0.0	nan	0.0570	0.2288	0.3788	0.0	0.4990	0.0927	0.1308	0.2841	0.0	0.0	0.0	0.0	0.0465	0.1502	0.3199
1.5653	22.5	900	1.7222	0.1272	0.2142	0.4277	nan	0.6924	0.1003	0.1794	0.5018	0.0	0.0	0.0	nan	0.0568	0.2295	0.3814	0.0	0.4969	0.0914	0.1341	0.2837	0.0	0.0	0.0	0.0	0.0466	0.1523	0.3209
1.5196	25.0	1000	1.7138	0.1266	0.2136	0.4273	nan	0.6939	0.0982	0.1706	0.5041	0.0	0.0	0.0	nan	0.0557	0.2283	0.3849	0.0	0.4965	0.0899	0.1288	0.2845	0.0	0.0	0.0	0.0	0.0462	0.1513	0.3225

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0