import os
import cv2
import torch
from model import U2NET
from torch.autograd import Variable
import numpy as np
from huggingface_hub import hf_hub_download
import gradio as gr

class PortraitGenerator:
    def __init__(self):
        self.u2net = self.load_u2net_model()

    def normPRED(self, d):
        return (d - torch.min(d)) / (torch.max(d) - torch.min(d))

    def inference(self, input_img):
        input_img = input_img / np.max(input_img)
        tmpImg = np.zeros((input_img.shape[0], input_img.shape[1], 3))
        tmpImg[:, :, 0] = (input_img[:, :, 2] - 0.406) / 0.225
        tmpImg[:, :, 1] = (input_img[:, :, 1] - 0.456) / 0.224
        tmpImg[:, :, 2] = (input_img[:, :, 0] - 0.485) / 0.229
        tmpImg = torch.from_numpy(tmpImg.transpose((2, 0, 1))[np.newaxis, :, :, :]).type(torch.FloatTensor)
        tmpImg = Variable(tmpImg.cuda() if torch.cuda.is_available() else tmpImg)
        d1, _, _, _, _, _, _ = self.u2net(tmpImg)
        pred = self.normPRED(1.0 - d1[:, 0, :, :])
        return pred.cpu().data.numpy().squeeze()

    def adjust_image(self, img, apply_bw, brightness, contrast, saturation, white_balance, hue, highlights_shadows, sharpness, noise_reduction):
        # Convert to grayscale if needed
        if apply_bw:
            img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
            img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

        # Adjust brightness and contrast
        img = cv2.convertScaleAbs(img, alpha=contrast / 50.0, beta=brightness - 50)

        # Adjust saturation
        if saturation != 50:
            hsv_img = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
            hsv_img[:, :, 1] = np.clip(hsv_img[:, :, 1] * (saturation / 50.0), 0, 255)
            img = cv2.cvtColor(hsv_img, cv2.COLOR_HSV2BGR)

        # Adjust white balance
        if white_balance != 50:
            img = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
            l, a, b = cv2.split(img)
            a = a * (white_balance / 50.0)
            b = b * (white_balance / 50.0)
            img = cv2.merge((l, a, b))
            img = cv2.cvtColor(img, cv2.COLOR_LAB2BGR)

        # Adjust hue
        if hue != 50:
            hsv_img = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
            hsv_img[:, :, 0] = np.clip(hsv_img[:, :, 0] * (hue / 50.0), 0, 180)
            img = cv2.cvtColor(hsv_img, cv2.COLOR_HSV2BGR)

        # Adjust highlights and shadows
        if highlights_shadows != 50:
            img = cv2.convertScaleAbs(img, alpha=1.0, beta=(highlights_shadows - 50) * 5.1)

        # Adjust sharpness
        if sharpness != 50:
            kernel = np.array([[-1, -1, -1], [-1, 9, -1], [-1, -1, -1]]) * (sharpness / 50.0)
            img = cv2.filter2D(img, -1, kernel)

        # Reduce noise
        if noise_reduction != 50:
            img = cv2.fastNlMeansDenoisingColored(img, None, noise_reduction / 50.0 * 10, noise_reduction / 50.0 * 10, 7, 21)

        return img

    def process_image(self, img, apply_bw, brightness, contrast, saturation, white_balance, hue, highlights_shadows, sharpness, noise_reduction, apply_adjustments, generate_final):
        if not generate_final:
            preview_img = self.adjust_image(img, apply_bw, brightness, contrast, saturation, white_balance, hue, highlights_shadows, sharpness, noise_reduction)
            return preview_img
        
        adjusted_img = self.adjust_image(img, apply_bw, brightness, contrast, saturation, white_balance, hue, highlights_shadows, sharpness, noise_reduction)
        result = self.inference(adjusted_img)
        return (result * 255).astype(np.uint8)

    def load_u2net_model(self):
        model_path = hf_hub_download(repo_id="Arrcttacsrks/U2net", filename="u2net_portrait.pth", use_auth_token=os.getenv("HF_TOKEN"))
        net = U2NET(3, 1)
        net.load_state_dict(torch.load(model_path, map_location="cuda" if torch.cuda.is_available() else "cpu"))
        net.eval()
        return net

def main():
    portrait_generator = PortraitGenerator()

    iface = gr.Interface(
        fn=portrait_generator.process_image,
        inputs=[
            gr.Image(type="numpy", label="Upload your image"),
            gr.Checkbox(label="Black & White Image"),
            gr.Slider(0, 100, value=50, label="Brightness"),
            gr.Slider(0, 100, value=50, label="Contrast"),
            gr.Slider(0, 100, value=50, label="Saturation"),
            gr.Slider(0, 100, value=50, label="White Balance"),
            gr.Slider(0, 100, value=50, label="Hue"),
            gr.Slider(0, 100, value=50, label="Highlights and Shadows"),
            gr.Slider(0, 100, value=50, label="Sharpness"),
            gr.Slider(0, 100, value=50, label="Noise Reduction"),
            gr.Checkbox(label="Apply Adjustments"),
            gr.Checkbox(label="Generate Final Portrait")
        ],
        outputs=gr.Image(type="numpy", label="Preview or Portrait Result"),
        title="Portrait Generation with U2NET",
        description="Upload an image to generate its portrait with optional adjustments. Enable 'Generate Final Portrait' for final output."
    )

    iface.launch()

if __name__ == "__main__":
    main()