Download app.py from Nvd/XXXX: direct link, hf CLI and curl.
- Browser
- Download file 1.16 kB
-
https://huggingface.co/spaces/Nvd/XXXX/resolve/main/app.py
- Command line
-
hf download hf://spaces/Nvd/XXXX/app.py
-
curl -L -o app.py https://huggingface.co/spaces/Nvd/XXXX/resolve/main/app.py
1.16 kB
| import torch | |
| import torchvision.transforms as transforms | |
| import gradio as gr | |
| from PIL import Image | |
| from model import SimpleCNN | |
| def preprocess_image(image): | |
| transform = transforms.Compose([ | |
| transforms.Resize((32, 32)), | |
| transforms.ToTensor(), | |
| transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), | |
| ]) | |
| image = Image.fromarray(image) | |
| # Ensure that the image has three channels (RGB) | |
| if image.mode != 'RGB': | |
| image = image.convert('RGB') | |
| image = transform(image) | |
| image = image.unsqueeze(0) | |
| return image | |
| def predict_image(model, image): | |
| with torch.no_grad(): | |
| output = model(image) | |
| _, predicted = torch.max(output.data, 1) | |
| return predicted.item() | |
| def main(): | |
| model = SimpleCNN() | |
| model.load_state_dict(torch.load('cifar10_model.pth')) | |
| # Set the model to evaluation mode | |
| model.eval() | |
| iface = gr.Interface( | |
| fn=lambda img: predict_image(model, preprocess_image(img)), | |
| inputs=gr.Image(), | |
| outputs="label", | |
| live=True, | |
| ) | |
| iface.launch() | |
| if __name__ == "__main__": | |
| main() |