segformer-b5-finetuned-Hiking

This model is a fine-tuned version of nvidia/mit-b5 on the twdent/Hiking dataset. It achieves the following results on the evaluation set:

• Loss: 0.1401
• Mean Iou: 0.6237
• Mean Accuracy: 0.9673
• Overall Accuracy: 0.9683
• Accuracy Unlabeled: nan
• Accuracy Traversable: 0.9641
• Accuracy Non-traversable: 0.9705
• Iou Unlabeled: 0.0
• Iou Traversable: 0.9178
• Iou Non-traversable: 0.9532

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
0.2675	1.33	20	0.2742	0.6058	0.9616	0.9550	nan	0.9826	0.9406	0.0	0.8853	0.9321
0.1418	2.67	40	0.1827	0.6073	0.9562	0.9566	nan	0.9548	0.9575	0.0	0.8858	0.9360
0.0949	4.0	60	0.1561	0.6002	0.9382	0.9523	nan	0.8931	0.9832	0.0	0.8692	0.9314
0.0684	5.33	80	0.1364	0.6135	0.9556	0.9614	nan	0.9369	0.9742	0.0	0.8967	0.9437
0.0627	6.67	100	0.1289	0.6122	0.9506	0.9610	nan	0.9177	0.9836	0.0	0.8928	0.9438
0.0625	8.0	120	0.1097	0.6208	0.9610	0.9658	nan	0.9458	0.9762	0.0	0.9113	0.9510
0.0371	9.33	140	0.1361	0.6130	0.9551	0.9610	nan	0.9361	0.9741	0.0	0.8959	0.9431
0.0409	10.67	160	0.1239	0.6194	0.9615	0.9653	nan	0.9494	0.9737	0.0	0.9086	0.9494
0.0457	12.0	180	0.0993	0.6281	0.9715	0.9713	nan	0.9723	0.9707	0.0	0.9264	0.9579
0.0368	13.33	200	0.1354	0.6146	0.9563	0.9617	nan	0.9389	0.9737	0.0	0.8993	0.9446
0.0667	14.67	220	0.1208	0.6171	0.9565	0.9644	nan	0.9316	0.9815	0.0	0.9032	0.9482
0.029	16.0	240	0.0946	0.6291	0.9695	0.9724	nan	0.9606	0.9785	0.0	0.9276	0.9596
0.0467	17.33	260	0.1188	0.6224	0.9655	0.9676	nan	0.9589	0.9721	0.0	0.9151	0.9522
0.0449	18.67	280	0.1201	0.6212	0.9638	0.9667	nan	0.9545	0.9731	0.0	0.9125	0.9511
0.0353	20.0	300	0.1285	0.6234	0.9687	0.9681	nan	0.9706	0.9668	0.0	0.9174	0.9527
0.025	21.33	320	0.1292	0.6204	0.9641	0.9659	nan	0.9582	0.9699	0.0	0.9114	0.9500
0.0244	22.67	340	0.1352	0.6208	0.9665	0.9664	nan	0.9667	0.9662	0.0	0.9124	0.9501
0.035	24.0	360	0.1260	0.6252	0.9699	0.9693	nan	0.9718	0.9681	0.0	0.9211	0.9544
0.0295	25.33	380	0.1190	0.6244	0.9669	0.9688	nan	0.9607	0.9730	0.0	0.9190	0.9543
0.032	26.67	400	0.1258	0.6253	0.9694	0.9695	nan	0.9693	0.9695	0.0	0.9211	0.9547
0.0241	28.0	420	0.1255	0.6230	0.9658	0.9678	nan	0.9593	0.9723	0.0	0.9164	0.9527
0.0246	29.33	440	0.1273	0.6238	0.9675	0.9683	nan	0.9651	0.9699	0.0	0.9179	0.9534
0.0214	30.67	460	0.1321	0.6233	0.9670	0.9675	nan	0.9652	0.9687	0.0	0.9171	0.9527
0.0236	32.0	480	0.1289	0.6241	0.9687	0.9685	nan	0.9695	0.9679	0.0	0.9189	0.9534
0.0238	33.33	500	0.1309	0.6234	0.9664	0.9680	nan	0.9612	0.9716	0.0	0.9172	0.9529
0.0204	34.67	520	0.1271	0.6249	0.9681	0.9693	nan	0.9643	0.9719	0.0	0.9201	0.9547
0.0243	36.0	540	0.1264	0.6248	0.9679	0.9693	nan	0.9636	0.9723	0.0	0.9196	0.9547
0.0259	37.33	560	0.1305	0.6226	0.9656	0.9679	nan	0.9582	0.9730	0.0	0.9154	0.9525
0.0341	38.67	580	0.1277	0.6245	0.9674	0.9690	nan	0.9623	0.9725	0.0	0.9192	0.9543
0.0275	40.0	600	0.1369	0.6221	0.9653	0.9672	nan	0.9590	0.9715	0.0	0.9147	0.9516
0.0303	41.33	620	0.1380	0.6235	0.9674	0.9681	nan	0.9650	0.9698	0.0	0.9175	0.9530
0.0207	42.67	640	0.1389	0.6237	0.9677	0.9682	nan	0.9662	0.9692	0.0	0.9180	0.9531
0.0231	44.0	660	0.1369	0.6243	0.9679	0.9688	nan	0.9652	0.9707	0.0	0.9190	0.9538
0.0249	45.33	680	0.1379	0.6237	0.9672	0.9683	nan	0.9640	0.9705	0.0	0.9179	0.9532
0.0382	46.67	700	0.1384	0.6239	0.9677	0.9685	nan	0.9650	0.9704	0.0	0.9182	0.9534
0.0238	48.0	720	0.1420	0.6230	0.9668	0.9677	nan	0.9640	0.9697	0.0	0.9166	0.9524
0.0212	49.33	740	0.1401	0.6237	0.9673	0.9683	nan	0.9641	0.9705	0.0	0.9178	0.9532

Framework versions

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