ViT-threat-classification-v2

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on an unknown dataset. This is model created as a prrof of concept for a Carleton University computer vision event. It is by no means meant to be used in deliverable systems in its current state, and should be used exclusively for research and development. It achieves the following results on the evaluation set:

• Loss: 0.0381
• F1: 0.9657
• Precision: 0.9563
• Recall: 0.9752

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Precision	Recall
0.0744	0.9985	326	0.0576	0.9466	0.9738	0.9208
0.0449	2.0	653	0.0397	0.9641	0.9747	0.9538
0.0207	2.9985	979	0.0409	0.9647	0.9607	0.9686
0.0342	4.0	1306	0.0382	0.9650	0.9518	0.9785
0.0286	4.9923	1630	0.0381	0.9657	0.9563	0.9752

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3