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

from tensorflow.keras import models 
from tensorflow.keras.preprocessing.image import load_img
import tensorflow as tf
from hugsvision.inference.TorchVisionClassifierInference import TorchVisionClassifierInference

models_name = [
    "VGG16",
    "DenseNet121",
    "DenseNet"
]

# open categories.txt in read mode
categories = open("categories.txt", "r")
labels = categories.readline().split(";")

# create a radio
radio = gr.inputs.Radio(models_name, default="DenseNet121", type="value")

def predict_image(image, model_name):
    # model create by keras 
    if model_name == "DenseNet":
        image = np.array(image) / 255
        image = np.expand_dims(image, axis=0)

        model = model = models.load_model("./models/" + model_name + "/model.h5")
        pred = model.predict(image)

        pred = dict((labels[i], "%.2f" % pred[0][i]) for i in range(len(labels)))
        
    # model create by HugsVision
    else:
        image = Image.fromarray(np.uint8(image)).convert('RGB')
        classifier = TorchVisionClassifierInference(
            model_path = "./models/" + model_name
        )

        pred = classifier.predict_image(img=image, return_str=False)

        for key in pred.keys():
            pred[key] = pred[key]/100
    

    print(pred)
    return pred

image = gr.inputs.Image(shape=(300, 300), label="Upload Your Image Here")
label = gr.outputs.Label(num_top_classes=len(labels))

samples = [["samples/" + p + ".jpg"] for p in labels]
        
interface = gr.Interface(
    fn=predict_image, 
    inputs=[image, radio], 
    outputs=label, 
    capture_session=True, 
    allow_flagging=False,
    title="🦈 Shark image classifier",
    description="Made with HugsVision & ❤️",
    examples=samples,
    theme=None
) 
interface.launch()