import gradio as gr
from urllib.request import urlopen
from PIL import Image
import timm
import torch
import time

model = timm.create_model("hf_hub:Marqo/nsfw-image-detection-384", pretrained=True)
model = model.eval()

data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

def predict(image):
    start_time = time.time()
    with torch.no_grad():
        input_tensor = transforms(image).unsqueeze(0)
        output = model(input_tensor).softmax(dim=-1).cpu()
        class_names = model.pretrained_cfg["label_names"]
        result = {class_names[i]: float(output[0, i]) for i in range(len(class_names))}
    end_time = time.time()
    inference_time = end_time - start_time
    return result, f"Inference time: {inference_time:.2f} seconds"


demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", height=512),
    outputs=[
        gr.Label(num_top_classes=2),
        gr.Textbox(label="Inference Time")
    ],
    title="NSFW Image Detection",
    description=(
        "Upload an image to detect if it is **NSFW (Not Safe For Work)** or **Safe For Work (SFW)**.\n\n"
        "This app uses the [Marqo/nsfw-image-detection-384](https://huggingface.co/Marqo/nsfw-image-detection-384) "
        "image classification model from Hugging Face's `timm` library."
    )
)

if __name__ == "__main__":
    demo.launch()