segformer-b1-finetuned-Hiking

This model is a fine-tuned version of nvidia/mit-b1 on the twdent/Hiking dataset. It achieves the following results on the evaluation set:

• Loss: 0.1129
• Mean Iou: 0.6261
• Mean Accuracy: 0.9707
• Overall Accuracy: 0.9700
• Accuracy Unlabeled: nan
• Accuracy Traversable: 0.9730
• Accuracy Non-traversable: 0.9684
• Iou Unlabeled: 0.0
• Iou Traversable: 0.9226
• Iou Non-traversable: 0.9557

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Traversable	Accuracy Non-traversable	Iou Unlabeled	Iou Traversable	Iou Non-traversable
0.5526	1.33	20	0.7232	0.5814	0.9404	0.9333	nan	0.9630	0.9177	0.0	0.8433	0.9009
0.3768	2.67	40	0.3616	0.5929	0.9504	0.9453	nan	0.9666	0.9341	0.0	0.8605	0.9183
0.3271	4.0	60	0.2683	0.6080	0.9606	0.9571	nan	0.9718	0.9494	0.0	0.8883	0.9358
0.2377	5.33	80	0.2754	0.5869	0.9256	0.9429	nan	0.8706	0.9807	0.0	0.8415	0.9191
0.2242	6.67	100	0.2299	0.6027	0.9447	0.9545	nan	0.9134	0.9760	0.0	0.8739	0.9341
0.2458	8.0	120	0.1939	0.6207	0.9604	0.9667	nan	0.9402	0.9805	0.0	0.9107	0.9513
0.1541	9.33	140	0.1988	0.6121	0.9654	0.9599	nan	0.9831	0.9477	0.0	0.8967	0.9395
0.1448	10.67	160	0.1722	0.6202	0.9677	0.9662	nan	0.9725	0.9629	0.0	0.9112	0.9494
0.1533	12.0	180	0.2112	0.5951	0.9570	0.9454	nan	0.9941	0.9199	0.0	0.8676	0.9176
0.107	13.33	200	0.1658	0.6139	0.9626	0.9614	nan	0.9665	0.9587	0.0	0.8992	0.9426
0.109	14.67	220	0.1342	0.6267	0.9714	0.9712	nan	0.9724	0.9705	0.0	0.9231	0.9569
0.1092	16.0	240	0.1448	0.6173	0.9690	0.9636	nan	0.9860	0.9519	0.0	0.9065	0.9453
0.0971	17.33	260	0.1282	0.6216	0.9691	0.9673	nan	0.9747	0.9635	0.0	0.9136	0.9512
0.1448	18.67	280	0.1504	0.6155	0.9661	0.9619	nan	0.9795	0.9526	0.0	0.9032	0.9434
0.0797	20.0	300	0.1312	0.6209	0.9669	0.9666	nan	0.9680	0.9659	0.0	0.9124	0.9503
0.0766	21.33	320	0.1164	0.6251	0.9667	0.9696	nan	0.9574	0.9760	0.0	0.9198	0.9555
0.0822	22.67	340	0.1365	0.6171	0.9638	0.9639	nan	0.9635	0.9641	0.0	0.9050	0.9464
0.075	24.0	360	0.1401	0.6160	0.9679	0.9621	nan	0.9862	0.9495	0.0	0.9046	0.9433
0.0684	25.33	380	0.1317	0.6190	0.9687	0.9645	nan	0.9822	0.9552	0.0	0.9099	0.9472
0.0767	26.67	400	0.1293	0.6195	0.9699	0.9651	nan	0.9851	0.9547	0.0	0.9107	0.9478
0.0576	28.0	420	0.1195	0.6236	0.9701	0.9679	nan	0.9771	0.9631	0.0	0.9180	0.9529
0.0596	29.33	440	0.1179	0.6248	0.9717	0.9692	nan	0.9794	0.9639	0.0	0.9204	0.9541
0.0564	30.67	460	0.1110	0.6264	0.9719	0.9701	nan	0.9777	0.9661	0.0	0.9233	0.9559
0.0496	32.0	480	0.1063	0.6284	0.9726	0.9714	nan	0.9761	0.9690	0.0	0.9271	0.9581
0.0722	33.33	500	0.1073	0.6272	0.9712	0.9711	nan	0.9716	0.9708	0.0	0.9244	0.9573
0.0465	34.67	520	0.1228	0.6220	0.9692	0.9669	nan	0.9763	0.9620	0.0	0.9150	0.9510
0.0655	36.0	540	0.1142	0.6245	0.9704	0.9689	nan	0.9752	0.9656	0.0	0.9196	0.9540
0.0516	37.33	560	0.1197	0.6238	0.9687	0.9684	nan	0.9696	0.9677	0.0	0.9181	0.9533
0.0774	38.67	580	0.1114	0.6266	0.9706	0.9704	nan	0.9712	0.9700	0.0	0.9234	0.9565
0.0572	40.0	600	0.1124	0.6261	0.9707	0.9700	nan	0.9730	0.9684	0.0	0.9227	0.9558
0.0554	41.33	620	0.1116	0.6273	0.9718	0.9709	nan	0.9747	0.9688	0.0	0.9248	0.9570
0.0438	42.67	640	0.1192	0.6259	0.9707	0.9694	nan	0.9746	0.9667	0.0	0.9225	0.9551
0.0486	44.0	660	0.1186	0.6248	0.9709	0.9689	nan	0.9775	0.9643	0.0	0.9203	0.9540
0.0582	45.33	680	0.1194	0.6250	0.9705	0.9691	nan	0.9751	0.9660	0.0	0.9209	0.9542
0.0643	46.67	700	0.1157	0.6252	0.9706	0.9692	nan	0.9750	0.9662	0.0	0.9209	0.9546
0.057	48.0	720	0.1181	0.6251	0.9708	0.9691	nan	0.9763	0.9654	0.0	0.9210	0.9543
0.0468	49.33	740	0.1129	0.6261	0.9707	0.9700	nan	0.9730	0.9684	0.0	0.9226	0.9557

Framework versions

• Transformers 4.35.0.dev0
• Pytorch 2.1.0+cu118
• Datasets 2.14.5
• Tokenizers 0.14.0