vehicle_classification

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

• Loss: 0.5738
• Accuracy: 0.8467

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	147	1.4917	0.7785
No log	2.0	294	1.0285	0.8160
No log	3.0	441	0.8369	0.8177
1.294	4.0	588	0.7112	0.8399
1.294	5.0	735	0.6621	0.8313
1.294	6.0	882	0.5977	0.8450
0.4624	7.0	1029	0.5856	0.8518
0.4624	8.0	1176	0.6511	0.8160
0.4624	9.0	1323	0.6450	0.8365
0.4624	10.0	1470	0.6241	0.8296
0.2619	11.0	1617	0.6217	0.8382
0.2619	12.0	1764	0.6504	0.8177
0.2619	13.0	1911	0.5994	0.8433
0.1776	14.0	2058	0.5969	0.8433
0.1776	15.0	2205	0.5693	0.8569

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2