|
|
--- |
|
|
license: mit |
|
|
--- |
|
|
|
|
|
# ResNetWildFireModel for Wildfire Classification |
|
|
|
|
|
## Model Details |
|
|
|
|
|
- **Model Architecture:** ResNet-18 (Modified) |
|
|
- **Framework:** PyTorch |
|
|
- **Input Shape:** 3-channel RGB images |
|
|
- **Number of Parameters:** ~11.7M (Based on ResNet-18) |
|
|
- **Output:** Binary classification (wildfire presence) |
|
|
|
|
|
## Model Description |
|
|
|
|
|
This model is a **fine-tuned ResNet-18** for wildfire classification. The pretrained **ResNet-18** backbone is used with its feature extractor **frozen**, while only the **final fully connected layer** is trained. The last fully connected layer has been replaced with a **single output neuron** for binary classification, predicting the presence of wildfire. |
|
|
|
|
|
## Training Details |
|
|
|
|
|
- **Optimizer:** Adam |
|
|
- **Batch Size:** 32 |
|
|
- **Loss Function:** Binary Cross-Entropy (BCE) |
|
|
- **Number of Epochs:** 10 |
|
|
- **Dataset:** [Wildfire Detection Image Data](https://www.kaggle.com/datasets/brsdincer/wildfire-detection-image-data) |
|
|
|
|
|
### Losses Per Epoch |
|
|
|
|
|
| Epoch | Training Loss | Validation Loss | |
|
|
|-------|---------------|-----------------| |
|
|
| 1 | 0.2182 | 0.0593 | |
|
|
| 2 | 0.0483 | 0.0508 | |
|
|
| 3 | 0.0347 | 0.0482 | |
|
|
| 4 | 0.0275 | 0.0461 | |
|
|
| 5 | 0.0253 | 0.0474 | |
|
|
| 6 | 0.0187 | 0.0457 | |
|
|
| 7 | 0.0131 | 0.0456 | |
|
|
| 8 | 0.0111 | 0.0451 | |
|
|
| 9 | 0.0096 | 0.0463 | |
|
|
| 10 | 0.0079 | 0.0474 | |
|
|
|
|
|
## License |
|
|
|
|
|
This model is released under the **MIT License**. |
|
|
|
|
|
--- |
|
|
|
|
|
|
