segformer-b1-finetuned-HikingHD

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

• Loss: 0.1067
• Mean Iou: 0.9379
• Mean Accuracy: 0.9665
• Overall Accuracy: 0.9684
• Accuracy Unlabeled: nan
• Accuracy Traversable: 0.9485
• Accuracy Non-traversable: 0.9845
• Iou Unlabeled: nan
• Iou Traversable: 0.9305
• Iou Non-traversable: 0.9452
• Local Tests:
  • Average inference time: 0.2622481801774767

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
0.3796	1.67	20	0.5835	0.6174	0.9605	0.9617	nan	0.9488	0.9721	0.0	0.9180	0.9343
0.3086	3.33	40	0.2597	0.9230	0.9589	0.9605	nan	0.9439	0.9739	nan	0.9143	0.9318
0.2717	5.0	60	0.2202	0.9386	0.9681	0.9687	nan	0.9626	0.9736	nan	0.9321	0.9451
0.2655	6.67	80	0.2127	0.9334	0.9658	0.9659	nan	0.9646	0.9670	nan	0.9267	0.9402
0.1603	8.33	100	0.1699	0.9383	0.9677	0.9686	nan	0.9601	0.9753	nan	0.9316	0.9450
0.2	10.0	120	0.1692	0.9289	0.9609	0.9637	nan	0.9342	0.9876	nan	0.9200	0.9378
0.1613	11.67	140	0.1389	0.9399	0.9676	0.9695	nan	0.9498	0.9853	nan	0.9328	0.9470
0.185	13.33	160	0.1612	0.9217	0.9566	0.9600	nan	0.9254	0.9878	nan	0.9116	0.9318
0.251	15.0	180	0.1461	0.9277	0.9603	0.9631	nan	0.9340	0.9865	nan	0.9187	0.9368
0.1038	16.67	200	0.1401	0.9248	0.9581	0.9616	nan	0.9258	0.9904	nan	0.9149	0.9346
0.0628	18.33	220	0.1556	0.9195	0.9548	0.9588	nan	0.9171	0.9924	nan	0.9086	0.9303
0.077	20.0	240	0.1439	0.9213	0.9561	0.9598	nan	0.9220	0.9902	nan	0.9110	0.9317
0.0714	21.67	260	0.1267	0.9344	0.9641	0.9666	nan	0.9404	0.9878	nan	0.9263	0.9425
0.081	23.33	280	0.1097	0.9397	0.9672	0.9694	nan	0.9470	0.9874	nan	0.9324	0.9470
0.09	25.0	300	0.1063	0.9402	0.9679	0.9696	nan	0.9522	0.9836	nan	0.9332	0.9472
0.0737	26.67	320	0.1045	0.9395	0.9674	0.9692	nan	0.9502	0.9845	nan	0.9323	0.9466
0.1173	28.33	340	0.1019	0.9427	0.9702	0.9708	nan	0.9644	0.9760	nan	0.9365	0.9488
0.0535	30.0	360	0.1132	0.9387	0.9674	0.9688	nan	0.9549	0.9799	nan	0.9317	0.9456
0.0693	31.67	380	0.1182	0.9340	0.9637	0.9664	nan	0.9389	0.9886	nan	0.9258	0.9422
0.0649	33.33	400	0.1108	0.9374	0.9662	0.9681	nan	0.9483	0.9841	nan	0.9300	0.9448
0.1581	35.0	420	0.1107	0.9368	0.9658	0.9678	nan	0.9473	0.9844	nan	0.9293	0.9443
0.0711	36.67	440	0.1011	0.9414	0.9690	0.9702	nan	0.9578	0.9801	nan	0.9348	0.9479
0.0743	38.33	460	0.1026	0.9400	0.9676	0.9695	nan	0.9500	0.9853	nan	0.9329	0.9471
0.0602	40.0	480	0.1029	0.9407	0.9681	0.9699	nan	0.9521	0.9841	nan	0.9337	0.9476
0.0768	41.67	500	0.1059	0.9386	0.9670	0.9688	nan	0.9502	0.9837	nan	0.9314	0.9458
0.0494	43.33	520	0.1076	0.9375	0.9663	0.9682	nan	0.9484	0.9842	nan	0.9302	0.9449
0.0359	45.0	540	0.1097	0.9369	0.9659	0.9679	nan	0.9473	0.9844	nan	0.9294	0.9444
0.0799	46.67	560	0.1070	0.9379	0.9666	0.9684	nan	0.9493	0.9838	nan	0.9306	0.9452
0.0685	48.33	580	0.1075	0.9378	0.9665	0.9684	nan	0.9489	0.9841	nan	0.9305	0.9452
0.0437	50.0	600	0.1067	0.9379	0.9665	0.9684	nan	0.9485	0.9845	nan	0.9305	0.9452

Framework versions

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