Luna-Skywalker's picture
Update README.md
984ae11 verified
metadata
library_name: transformers
license: apache-2.0
base_model: google/vit-base-patch16-224-in21k
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - precision
  - recall
  - f1
model-index:
  - name: crime_cctv_image_detection
    results: []

crime_cctv_image_detection

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

  • Loss: 0.0332
  • Accuracy: 0.9957
  • Precision: 0.9954
  • Recall: 0.9954
  • F1: 0.9954

Model description

This model was developed with the intention of effectively monitoring crime and making society a much safer place. This is a small part of the project that is under developement. I'd like to welcome you all to try this and please provide valuable feedback. I'll be uploading an example notebok for usage of this model very soon (as soon as I'm done with school lol)

Training and evaluation data

The total image count for the train subset is 1,266,345. The total image count for the test subset is 111,308.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-06
  • train_batch_size: 64
  • eval_batch_size: 16
  • seed: 42
  • 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_steps: 50
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1
0.7736 1.0 4608 0.1649 0.9874 0.9857 0.9827 0.9842
0.0836 2.0 9216 0.0487 0.9951 0.9948 0.9948 0.9948
0.0337 3.0 13824 0.0332 0.9957 0.9954 0.9954 0.9954

Framework versions

  • Transformers 4.53.3
  • Pytorch 2.6.0+cu124
  • Datasets 4.1.1
  • Tokenizers 0.21.2