date3k2/vit-real-fake-classification-v2

ViT Real Fake Image Classification

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

• Loss: 0.0398
• Accuracy: 0.9866
• F1: 0.9878
• Recall: 0.9854
• Precision: 0.9902

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Recall	Precision
0.1759	1.0	59	0.2212	0.9173	0.9229	0.8978	0.9495
0.1903	2.0	118	0.1047	0.9629	0.9659	0.9503	0.9819
0.0463	3.0	177	0.0824	0.9699	0.9730	0.9834	0.9628
0.0015	4.0	236	0.0763	0.9764	0.9787	0.9825	0.9749
0.0631	5.0	295	0.0794	0.9737	0.9759	0.9640	0.9880
0.0114	6.0	354	0.0582	0.9801	0.9819	0.9786	0.9853
0.0004	7.0	413	0.0662	0.9807	0.9824	0.9796	0.9853
0.0231	8.0	472	0.0713	0.9753	0.9773	0.9659	0.9890
0.0017	9.0	531	0.0518	0.9817	0.9834	0.9796	0.9872
0.0268	10.0	590	0.0385	0.9839	0.9855	0.9903	0.9807

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1