|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- image-classification |
|
|
- swin-transformer |
|
|
- swin-v2 |
|
|
- vision |
|
|
- gradients |
|
|
- jet-colormap |
|
|
--- |
|
|
|
|
|
# SwinV2-Large Classifier (384x384, Jet-Colored Gradients) |
|
|
|
|
|
This model is a **SwinV2-Large (384x384)** vision transformer trained for image classification on **Jet-colored gradient maps**. The model learns to identify visual patterns in synthetic or colormap-encoded data to be suitable for detecting GAN generated images. |
|
|
|
|
|
## 🧩 Model Details |
|
|
|
|
|
- **Architecture**: SwinV2-Large (384x384 input resolution) |
|
|
- **Framework**: PyTorch |
|
|
- **Training Data**: Jet-colored gradients (Mukhbir dataset on kaggle) |
|
|
- **Use case**: Classification of Real or Fake(Gan) |
|
|
|
|
|
## 🛠️ How to Use |
|
|
|
|
|
```python |
|
|
from huggingface_hub import hf_hub_download |
|
|
import torch |
|
|
|
|
|
# Download the model |
|
|
model_path = hf_hub_download(repo_id="mukhbiir/Swin_Classifier", filename="model.pt") |
|
|
model = torch.load(model_path) |
|
|
model.eval() |
|
|
``` |
|
|
|
|
|
|
