segformer-b0-finetuned-robot-hiking

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

• Loss: 0.1458
• Mean Iou: 0.6158
• Mean Accuracy: 0.9603
• Overall Accuracy: 0.9619
• Accuracy Unlabeled: nan
• Accuracy Traversable: 0.9546
• Accuracy Non-traversable: 0.9661
• Iou Unlabeled: 0.0
• Iou Traversable: 0.9044
• Iou Non-traversable: 0.9429

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
0.5363	1.33	20	0.7433	0.5581	0.9246	0.9153	nan	0.9573	0.8919	0.0	0.8030	0.8712
0.3689	2.67	40	0.4666	0.5567	0.9281	0.9137	nan	0.9791	0.8771	0.0	0.8030	0.8670
0.2706	4.0	60	0.3204	0.5880	0.9479	0.9407	nan	0.9732	0.9226	0.0	0.8550	0.9089
0.2338	5.33	80	0.2881	0.5985	0.9504	0.9497	nan	0.9527	0.9481	0.0	0.8719	0.9236
0.2068	6.67	100	0.2556	0.9022	0.9521	0.9521	nan	0.9522	0.9521	nan	0.8770	0.9273
0.1764	8.0	120	0.2401	0.6024	0.9539	0.9528	nan	0.9577	0.9500	0.0	0.8792	0.9280
0.2639	9.33	140	0.2588	0.5937	0.9504	0.9455	nan	0.9680	0.9329	0.0	0.8646	0.9166
0.1813	10.67	160	0.2124	0.6030	0.9526	0.9530	nan	0.9513	0.9540	0.0	0.8801	0.9287
0.1407	12.0	180	0.1938	0.6055	0.9518	0.9554	nan	0.9388	0.9647	0.0	0.8836	0.9328
0.13	13.33	200	0.1881	0.6062	0.9524	0.9558	nan	0.9403	0.9644	0.0	0.8854	0.9333
0.107	14.67	220	0.2092	0.5967	0.9530	0.9474	nan	0.9725	0.9334	0.0	0.8708	0.9194
0.1282	16.0	240	0.1803	0.6065	0.9536	0.9555	nan	0.9471	0.9602	0.0	0.8869	0.9328
0.146	17.33	260	0.1912	0.6028	0.9559	0.9519	nan	0.9700	0.9418	0.0	0.8814	0.9269
0.1011	18.67	280	0.1769	0.6079	0.9598	0.9561	nan	0.9727	0.9468	0.0	0.8907	0.9330
0.1124	20.0	300	0.1580	0.6135	0.9582	0.9608	nan	0.9491	0.9673	0.0	0.8995	0.9411
0.0801	21.33	320	0.1614	0.6113	0.9582	0.9588	nan	0.9563	0.9602	0.0	0.8960	0.9380
0.0831	22.67	340	0.1540	0.6130	0.9608	0.9601	nan	0.9633	0.9584	0.0	0.8994	0.9396
0.0599	24.0	360	0.1641	0.6098	0.9584	0.9576	nan	0.9614	0.9554	0.0	0.8935	0.9358
0.0955	25.33	380	0.1711	0.6084	0.9597	0.9562	nan	0.9720	0.9474	0.0	0.8917	0.9334
0.0667	26.67	400	0.1618	0.6109	0.9574	0.9583	nan	0.9543	0.9605	0.0	0.8954	0.9373
0.0783	28.0	420	0.1640	0.6089	0.9589	0.9568	nan	0.9665	0.9513	0.0	0.8924	0.9343
0.0743	29.33	440	0.1512	0.6145	0.9582	0.9612	nan	0.9478	0.9686	0.0	0.9016	0.9419
0.0775	30.67	460	0.1574	0.6131	0.9583	0.9598	nan	0.9528	0.9637	0.0	0.8995	0.9398
0.0773	32.0	480	0.1464	0.6157	0.9610	0.9621	nan	0.9573	0.9647	0.0	0.9043	0.9428
0.0575	33.33	500	0.1600	0.6085	0.9568	0.9564	nan	0.9583	0.9554	0.0	0.8912	0.9343
0.0729	34.67	520	0.1540	0.6105	0.9569	0.9577	nan	0.9541	0.9597	0.0	0.8946	0.9369
0.1409	36.0	540	0.1557	0.6112	0.9584	0.9586	nan	0.9575	0.9593	0.0	0.8962	0.9376
0.0543	37.33	560	0.1607	0.6103	0.9546	0.9581	nan	0.9422	0.9670	0.0	0.8938	0.9373
0.063	38.67	580	0.1622	0.6099	0.9558	0.9574	nan	0.9504	0.9613	0.0	0.8933	0.9364
0.0549	40.0	600	0.1543	0.6118	0.9571	0.9590	nan	0.9503	0.9639	0.0	0.8969	0.9386
0.0766	41.33	620	0.1481	0.6139	0.9575	0.9606	nan	0.9467	0.9684	0.0	0.9005	0.9412
0.0616	42.67	640	0.1485	0.6151	0.9600	0.9614	nan	0.9550	0.9649	0.0	0.9032	0.9422
0.0872	44.0	660	0.1511	0.6144	0.9607	0.9609	nan	0.9602	0.9612	0.0	0.9022	0.9409
0.0677	45.33	680	0.1510	0.6139	0.9599	0.9604	nan	0.9580	0.9618	0.0	0.9012	0.9405
0.1075	46.67	700	0.1506	0.6145	0.9606	0.9610	nan	0.9592	0.9619	0.0	0.9024	0.9411
0.0485	48.0	720	0.1450	0.6159	0.9595	0.9621	nan	0.9506	0.9685	0.0	0.9043	0.9433
0.0972	49.33	740	0.1458	0.6158	0.9603	0.9619	nan	0.9546	0.9661	0.0	0.9044	0.9429

Framework versions

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