import gradio as gr
import torch
from PIL import Image
import numpy as np

model = torch.hub.load('ultralytics/yolov5', 'custom', path='best.pt')

import json

def yolo(im, size=1024):
    g = (size / max(im.size))  # gain
    im = im.resize((int(x * g) for x in im.size), Image.LANCZOS)  # resize

    results = model(im)  # inference
    if results.imgs:
        writable_img = np.copy(results.imgs[0])  # Create a writable copy of the image data
        results.imgs[0] = writable_img
        results.render()  # updates results.imgs with boxes and labels

    df = results.pandas().xyxy[0].to_json(orient="records")
    res = json.loads(df)

    return [
        Image.fromarray(results.imgs[0]),
        res
    ]


inputs = gr.Image(type='pil', label="Original Image")
outputs = [
    gr.Image(type="pil", label="Output Image"), 
    gr.JSON()
]

title = "YOLOv5 NDL-DocL Datasets"
description = "YOLOv5 NDL-DocL Datasets Gradio demo for object detection. Upload an image or click an example image to use."
article = "<p style='text-align: center'>YOLOv5 NDL-DocL Datasets is an object detection model trained on the <a href=\"https://github.com/ndl-lab/layout-dataset\">NDL-DocL Datasets</a>.</p>"

# <a href=\"aaa\"></a>

examples = [
    ['『源氏物語』(東京大学総合図書館所蔵).jpg'],
    ['『源氏物語』(京都大学所蔵).jpg'],
    ['『平家物語』(国文学研究資料館提供).jpg']
]
demo = gr.Interface(yolo, inputs, outputs, title=title, description=description, article=article, examples=examples)

demo.launch(share=False)