OCR/app.py

from flask import Flask, render_template, request
from utils.util import *
from utils.conf import MAX_CONTENT_LENGTH
import numpy as np
import cv2
from paddleocr import PaddleOCR

# from paddleocr import PaddleOCR
# from utils import Args
# from paddleocr.tools.infer.predict_system import TextSystem
#
app = Flask(__name__)
app.config["JSON_AS_ASCII"] = False
# 待优化为 TextSystem
engine = PaddleOCR(
    use_gpu=False,
    enable_mkldnn=True,
    det_model_dir="models/det/",
    rec_model_dir="models/rec/",
    cls_model_dir="models/cls/",
    use_angle_cls=True,
    use_space_char=True
)


# args = Args(
#     use_gpu=False,
#     enable_mkldnn=True,
#     det_model_dir="models/det/",
#     rec_model_dir="models/rec/",
# )
# eng2 = TextSystem(args)
# warmup


@app.route("/")
def index():
    return render_template("index.html")


@app.route("/ocr-raw-api", methods=["POST"])
def ocr_raw():
    # 文件处理
    pic = request.files.get("picture")
    if pic is None:
        return Response("empty body")
    ext = get_ext_name(pic.filename)
    if not is_image_ext(ext):
        return Response("文件类型错误")
    content = pic.read()
    if len(content) > MAX_CONTENT_LENGTH:
        return Response("文件过大，请压缩后尝试")
    path = f"static/images/{current_time()}_{rand_str()}.{ext}"
    with open(path, "wb") as fp:
        fp.write(content)
        fp.close()

    # 内容识别
    array = cv2.imdecode(np.fromstring(content, np.uint8), 1)  # noqa
    ocr_res = engine.ocr(array)[0]
    res = [{"pos": it[0], "word": it[1][0], "rate": it[1][1]} for it in ocr_res]
    return Response(data=res)


@app.route("/ocr-filter", methods=["POST"])
def ocr_filter():
    # 文件处理
    pic = request.files.get("picture")
    if pic is None:
        return Response("empty body")
    ext = get_ext_name(pic.filename)
    if not is_image_ext(ext):
        return Response("文件类型错误")
    content = pic.read()
    if len(content) > MAX_CONTENT_LENGTH:
        return Response("文件过大，请压缩后尝试")
    path = f"static/images/{current_time()}_{rand_str()}.{ext}"
    with open(path, "wb") as fp:
        fp.write(content)
        fp.close()

    # 内容识别
    array = cv2.imdecode(np.fromstring(content, np.uint8), 1)  # noqa
    ocr_res = engine.ocr(array)[0]

    # 过滤出想要的数据
    res = [it[1][0] for it in ocr_res]
    return Response(data=res)


@app.route("/ocr-html", methods=["POST"])
def ocr_html():
    # 文件处理
    pic = request.files.get("picture")
    if pic is None:
        return Response("empty body")
    ext = get_ext_name(pic.filename)
    if not is_image_ext(ext):
        return Response("文件类型错误")
    content = pic.read()
    if len(content) > MAX_CONTENT_LENGTH:
        return Response("文件过大，请压缩后尝试")
    cur, rnd = current_time(), rand_str()
    raw_path = f"static/images/{cur}_{rnd}.{ext}"
    rec_path = f"static/images/{cur}_{rnd}-rec.{ext}"
    with open(raw_path, "wb") as fp:
        fp.write(content)
        fp.close()

    # 内容识别
    array = cv2.imdecode(np.fromstring(content, np.uint8), 1)  # noqa
    ocr_res = engine.ocr(array)[0]
    res = [{"pos": it[0], "word": it[1][0], "rate": it[1][1]} for it in ocr_res]

    # 画图
    rec_img_data = draw_img(array.shape, res)
    cv2.imwrite(rec_path, rec_img_data)  # noqa
    return render_template("result.html", raw=raw_path, rec=rec_path, data=res)


if __name__ == "__main__":
    app.run()

"""
step:
    pip install PaddlePaddle-GPU==2.4.2 PaddleOCR
    pip install PaddleHub
    hub install chinese_ocr_db_crnn_server


python E:/Project/Python/PaddleOCR/venv/Lib/site-packages/paddleocr/tools/infer/predict_system.py
    --image_dir="C:/Users/huimv/Pictures/Saved Pictures/"
    --det_model_dir="D:/BaiduNetdisk/module/aiengine/model/ocr/ch_pp-ocrv3_det_infer/"
    --rec_model_dir="D:/BaiduNetdisk/module/aiengine/model/ocr/ch_pp-ocrv3_rec_infer/"
    --cls_model_dir="D:/BaiduNetdisk/module/aiengine/model/ocr/ch_ppocr_mobile_v2.0_cls_infer/"
    --use_angle_cls=True --use_space_char=True --use_gpu=False

python tools/infer/predict_system.py
    --image_dir="E:/Project/Python/PaddleOCR/static/pic2.png"
    --det_model_dir="E:/Project/Python/PaddleOCR/models/det/"
    --rec_model_dir="E:/Project/Python/PaddleOCR/models/rec/"


deploy:
    python -m pip install PaddlePaddle==2.4.2 PaddleOCR -i https://pypi.tuna.tsinghua.edu.cn/simple
"""