import requests
import torch
from PIL import Image
from transformers import pipeline
from paddleocr import PaddleOCR
from transformers import BlipProcessor, BlipForConditionalGeneration


class translate_models:
    @staticmethod
    def translate_models(file_path, flag, pairs):

        # pairs = 1  # 默认为英转中
        # flag = 2  # flag：判断是进行文字识别还是图片理解  1:文字识别  2:图片理解
        # image_path_identify = 'D:\\py\\pythonProject\\image\\00056221.jpg'  # 用来文字识别的图片所在的文件路径
        # image_path_understand = r"C:\Users\16173\Desktop\ccf_cv\blue-throat-7077261__340.jpg"  # 用来进行图片理解的文件路径

        """事实证明除中文外，英语转成其他语言更难"""
        language_pairs = {
            "en-zh": "Helsinki-NLP/opus-mt-en-zh",  # 1:英语到中文
            "zh-en": "Helsinki-NLP/opus-mt-zh-en",  # 2:中文到英语
            "en-fr": "Helsinki-NLP/opus-mt-en-fr",  # 3:英语到法语
            "fr-en": "Helsinki-NLP/opus-mt-fr-en",  # 4:法语到英语
            "en-es": "Helsinki-NLP/opus-mt-en-es",  # 5:英语到西班牙语
            "es-en": "Helsinki-NLP/opus-mt-es-en",  # 6:西班牙语到英语
            "en-de": "Helsinki-NLP/opus-mt-en-de",  # 7:英语到德语
            "de-en": "Helsinki-NLP/opus-mt-de-en",  # 8:德语到英语
            "en-ru": "Helsinki-NLP/opus-mt-en-ru",  # 9:英语到俄语
            "ru-en": "Helsinki-NLP/opus-mt-ru-en",  # 10:俄语到英语
            "zh-fr": "Helsinki-NLP/opus-mt-zh-fr",  # 11:中文到法语
            "fr-zh": "Helsinki-NLP/opus-mt-fr-zh",  # 12:法语到中文
        }

        if flag == 1:
            image_path_identify = file_path# 用来文字识别的图片所在的文件路径

            if torch.cuda.is_available():
                # 使用 GPU
                ocr = PaddleOCR(use_angle_cls=True, lang='ch', use_gpu=True)
                print("使用 GPU 进行 OCR")
            else:
                # 使用 CPU
                ocr = PaddleOCR(use_angle_cls=True, lang='ch', use_gpu=False)
                print("使用 CPU 进行 OCR")
            flag = 0
            result = ocr.ocr(image_path_identify, cls=True)

            # 提取纯文字内容
            extracted_text = [res[1][0] for res in result[0]]

            """-----------------------------------------这是文字识别的输出--------------------------------------------------"""
            print("文字识别结果如下：")
            for text in extracted_text:
                print(text)

        if flag == 2:
            # 图片理解模块
            image_path_understand = file_path  # 用来进行图片理解的文件路径

            processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
            model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")

            image_path = image_path_understand  # 本地图片路径
            image = Image.open(image_path)

            # 处理图片
            inputs = processor(images=image, return_tensors="pt")
            output = model.generate(**inputs)

            # 解码生成的描述
            description = processor.decode(output[0], skip_special_tokens=True)
            print(description)

            if pairs == 1:
                language_pair = "en-zh"
            if pairs == 2:
                language_pair = "zh-en"
            if pairs == 3:
                language_pair = "en-fr"
            if pairs == 4:
                language_pair = "fr-en"
            if pairs == 5:
                language_pair = "en-es"
            if pairs == 6:
                language_pair = "es-en"
            if pairs == 7:
                language_pair = "en-de"
            if pairs == 8:
                language_pair = "de-en"
            if pairs == 9:
                language_pair = "en-ru"
            if pairs == 10:
                language_pair = "ru-en"
            if pairs == 11:
                language_pair = "zh-fr"
            if pairs == 12:
                language_pair = "fr-zh"
            translator = pipeline("translation", model=language_pairs[language_pair])

            # 翻译文本
            result = translator(description, max_length=40)
            translation_text = result[0]['translation_text']

            # 要把这里的translation_text传给前端/////
            print(translation_text)
            return translation_text