wildfire-classifier / README.md
arnaucas's picture
Update README.md
08cad7e
metadata
license: apache-2.0
base_model: google/vit-base-patch16-384
tags:
  - generated_from_trainer
  - climate
  - biology
metrics:
  - accuracy
model-index:
  - name: wildfire-classifier
    results: []
widget:
  - src: >-
      https://news.erau.edu/-/media/images/news/headlines/january-2023/wildfire-overhead-drone-shot.jpg?h=749&w=1000&hash=13476D2A9BBA829375B2EB7E83588E18
    example_title: Drone-shot
  - src: >-
      https://www.ecuadorforestofclouds.org/uploads/7/4/1/4/74143387/2015367_orig.jpg
    example_title: Cloudy forest

Wildfire classifier

This model is a fine-tuned version of google/vit-base-patch16-384 on the Kaggle Wildfire Dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2329
  • Accuracy: 0.9202

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0002
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 4

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.1208 1.28 100 0.2329 0.9202
0.0261 2.56 200 0.2469 0.9316
0.0007 3.85 300 0.2358 0.9392

Framework versions

  • Transformers 4.33.2
  • Pytorch 2.0.1+cu117
  • Datasets 2.14.5
  • Tokenizers 0.13.3

Aditional resources

Fine-tuning tutorial