dinov2_LoRA_Liveness_detection_v1.0

This model is a fine-tuned version of facebook/dinov2-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0073
• Accuracy: 0.9979
• F1: 0.9979
• Recall: 0.9979
• Precision: 0.9979

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: 512
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Recall	Precision
0.0256	0.4076	64	0.0211	0.9931	0.9931	0.9931	0.9931
0.0164	0.8153	128	0.0153	0.9943	0.9943	0.9943	0.9943
0.0073	1.2229	192	0.0116	0.9962	0.9962	0.9962	0.9962
0.0073	1.6306	256	0.0073	0.9974	0.9974	0.9974	0.9974
0.0082	2.0382	320	0.0076	0.9977	0.9977	0.9977	0.9977
0.0048	2.4459	384	0.0070	0.9977	0.9977	0.9977	0.9977
0.0019	2.8535	448	0.0073	0.9979	0.9979	0.9979	0.9979

Framework versions

• PEFT 0.14.0
• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1