Edit model card

Traffic level image classification

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

  • Loss: 0.4394
  • Accuracy: 0.8292
  • Precision: 0.8232
  • Recall: 0.7366
  • F1: 0.7721

Model description

Built from 6,000 images fetched from public traffic cameras in Norway to classify traffic levels from low, medium to high. Dataset is unbalanced skewed towards low traffic images.

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: 32
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1
0.6282 0.9843 47 0.5725 0.7644 0.7933 0.5918 0.6525
0.4486 1.9895 95 0.4630 0.8012 0.7964 0.6824 0.7213
0.3285 2.9948 143 0.4394 0.8292 0.8232 0.7366 0.7721
0.2391 4.0 191 0.4302 0.8115 0.7941 0.7333 0.7555
0.1814 4.9215 235 0.4365 0.8218 0.7993 0.7362 0.7631

Framework versions

  • Transformers 4.40.1
  • Pytorch 2.2.1+cu121
  • Datasets 2.19.0
  • Tokenizers 0.19.1
Downloads last month
63
Safetensors
Model size
85.8M params
Tensor type
F32
·

Finetuned from