vit-utility-poles

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the utility-poles-local dataset. It achieves the following results on the evaluation set:

• Loss: 4.1032
• Accuracy: 0.1818
• Precision: 0.1363
• Recall: 0.1818
• F1: 0.1347

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	1.0	20	4.4879	0.0260	0.0017	0.0260	0.0031
No log	2.0	40	4.4301	0.0390	0.0018	0.0390	0.0034
No log	3.0	60	4.3691	0.0649	0.0549	0.0649	0.0379
No log	4.0	80	4.3187	0.0779	0.1134	0.0779	0.0565
3.9392	5.0	100	4.2822	0.1169	0.1133	0.1169	0.0926
3.9392	6.0	120	4.2440	0.1299	0.1187	0.1299	0.0998
3.9392	7.0	140	4.2341	0.1169	0.1239	0.1169	0.1001
3.9392	8.0	160	4.2013	0.1558	0.1449	0.1558	0.1262
3.9392	9.0	180	4.1658	0.1688	0.1301	0.1688	0.1303
2.5046	10.0	200	4.1695	0.1429	0.1168	0.1429	0.1098
2.5046	11.0	220	4.1433	0.1688	0.1245	0.1688	0.1257
2.5046	12.0	240	4.1359	0.1818	0.1428	0.1818	0.1400
2.5046	13.0	260	4.1293	0.1688	0.1276	0.1688	0.1259
2.5046	14.0	280	4.1201	0.1688	0.1246	0.1688	0.1237
1.6736	15.0	300	4.1130	0.1818	0.1445	0.1818	0.1413
1.6736	16.0	320	4.1112	0.1818	0.1367	0.1818	0.1354
1.6736	17.0	340	4.1046	0.1688	0.1285	0.1688	0.1267
1.6736	18.0	360	4.1025	0.1818	0.1361	0.1818	0.1344
1.6736	19.0	380	4.1043	0.1818	0.1363	0.1818	0.1347
1.3141	20.0	400	4.1032	0.1818	0.1363	0.1818	0.1347

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu118
• Datasets 3.5.0
• Tokenizers 0.21.1