GPU 프로젝트 업로드

app.py

@@ -0,0 +1,251 @@
+from flask import Flask, request, redirect, url_for
+
+import os
+import time
+import json
+import shutil
+import datetime
+import pandas as pd
+
+from pathlib import Path
+from natsort import os_sorted
+
+from main import pointchecker
+from qna import categorize_qna
+from mul import detect_multiple
+from sub import detect_subjective
+from path import *
+from utils import *
+
+app = Flask(__name__)
+
+ALLOWED_FILE_EXTENSIONS = set(['pdf', 'png', 'jpg', 'jpeg'])
+
+def allowed_file(filename):
+    return '.' in filename and \
+           filename.rsplit('.', 1)[1] in ALLOWED_FILE_EXTENSIONS
+
+
+@app.route("/")
+def hello():
+    # id 생성 규칙 - 클라이언트 ip + 접속시간
+    id = getId()
+    return f'Hello World! <br><br> id : {id}'
+
+@app.route("/id/<id>", methods=["GET"])
+def get_json(client_id):
+    id_path = UPLOAD_FOLDER + "/" + client_id
+    json_path = id_path + "/" + "data.json"
+
+    with open(json_path) as f:
+        json_data = load(f)
+
+    return json_data, 200
+
+
+@app.route("/upload", methods=["POST"])
+def upload_files():
+    ## upload 폴더 생성 ##
+    try:
+        if not os.path.exists(UPLOAD_FOLDER):
+            os.mkdir(UPLOAD_FOLDER)
+    except:
+        pass
+    ## upload 폴더 생성 ##
+
+    if request.method == "POST":
+        files = request.files
+        datas = request.form
+        data = datas.get("data")
+        data = json.loads(data)
+
+        client_id = data["client_id"]
+        id_path = UPLOAD_FOLDER + "/" + client_id
+
+        print(client_id)
+        print(id_path)
+
+        ## id 폴더 생성 ##
+        try:
+            if not os.path.exists(id_path):
+                os.mkdir(id_path)
+        except:
+            pass
+        ## id 폴더 생성 ##
+
+        pdf = files["pdf"]
+        pdf_name = pdf.filename
+        pdf_path = os.path.join(id_path, pdf_name)
+        if pdf and allowed_file(pdf_name):
+            pdf.save(pdf_path)
+
+        test_name = data["test_name"]
+        copy_num = data["copy_num"]
+        total_qna_num = data["total_qna_num"]
+        testee_num = data["testee_num"]
+        test_category = data["test_category"]
+
+        print("파일 업로드 성공")
+
+    return redirect(url_for("plural_check",
+                            client_id=client_id, 
+                            test_name=test_name,
+                            copy_num=copy_num,
+                            total_qna_num=total_qna_num,
+                            testee_num=testee_num,
+                            test_category=test_category))
+
+
+# 다인용
+@app.route("/plural", methods=["GET"])
+def plural_check():
+    client_id = request.args.get("client_id", type=str)
+    test_name = request.args.get("test_name", type=int)
+    copy_num = request.args.get("copy_num", type=int)
+    total_qna_num = request.args.get("total_qna_num", type=int)
+    testee_num = request.args.get("testee_num", type=int)
+    test_category = request.args.getlist("test_category")
+
+    id_path = UPLOAD_FOLDER + "/" + client_id
+
+    start = time.time()
+    df = pd.DataFrame()
+    df = pointchecker(id_path, test_name, copy_num, total_qna_num, testee_num, test_category)
+    end = time.time()
+    point_eta = end - start
+
+    print()
+    print(df.set_index(keys=["testee_id", "file"], drop=True))
+    print()
+    print("point_eta: " + f"{point_eta:.2f} sec")
+
+    if len(df) == 0:
+        return "Error Occured", 200
+
+    json_data = df.to_json(orient="records")
+    json_path = id_path + "/" + "data.json"
+
+    with open(json_path, 'w') as f:
+        json.dump(json_data, f)
+
+    return json_data, 200
+
+
+# 1인용
+@app.route("/single")
+def single_check():
+    id = request.args.get("id", type=str)
+    id_path = UPLOAD_FOLDER + "/" + id
+    mul_path = id_path + "/mul"
+    sub_path = id_path + "/sub"
+
+    ## 결과 저장 폴더 생성 ##
+    makeFolder(id_path)
+    makeFolder(mul_path)
+    makeFolder(sub_path)
+    ## 결과 저장 폴더 생성 ##
+
+    print_intro()
+
+    original_pdf_file_path_list = []
+    original_pdf_file_path_list = os_sorted(Path(id_path).glob('*.pdf'))
+    convertPdfToJpg(original_pdf_file_path_list, id_path)
+
+    df = pd.DataFrame()
+    mul_df = pd.DataFrame()
+    sub_df = pd.DataFrame()
+
+    final_df = pd.DataFrame(columns=["testee_id", "file", "num", "testee_answer", "correct_answer"])
+
+    start = time.time()
+    categorize_qna(id_path)
+    end = time.time()
+    qna_eta = end - start
+
+    start = time.time()
+    mul_df = detect_multiple(id_path)
+    mul_df.sort_values(by=["num"], inplace=True)
+    end = time.time()
+    mul_eta = end - start
+    print()
+    print_full(mul_df)
+
+    start = time.time()
+    sub_df = detect_subjective(id_path)
+    sub_df.sort_values(by=["num"], inplace=True)
+    end = time.time()
+    sub_eta = end - start
+    print()
+    print_full(sub_df)
+
+    df = pd.concat([mul_df, sub_df], axis=0, ignore_index=True)
+    df.sort_values(by=["num"], inplace=True)
+    final_df = dfToFinalDf(df)
+    print()
+    print_full(final_df)
+
+    final_df.to_excel(excel_writer=id_path+"/df.xlsx")
+
+    print("qna_eta: " + f"{qna_eta:.2f} sec")
+    print("mul_eta: " + f"{mul_eta:.2f} sec")
+    print("sub_eta: " + f"{sub_eta:.2f} sec")
+
+    print_outro()
+
+    json_data = final_df.to_json(orient="records")
+    return json_data, 200
+
+
+# mul_test
+@app.route("/mul_test")
+def mul_check():
+    id = request.args.get("id", type=str)
+    id_path = UPLOAD_FOLDER + "/" + id
+    mul_path = id_path + "/mul"
+
+    ## 결과 저장 폴더 생성 ##
+    makeFolder(id_path)
+    makeFolder(mul_path)
+    ## 결과 저장 폴더 생성 ##
+
+    print_intro()
+
+    original_pdf_file_path_list = []
+    original_pdf_file_path_list = os_sorted(Path(id_path).glob('*.pdf'))
+    convertPdfToJpg(original_pdf_file_path_list, id_path)
+
+    df = pd.DataFrame()
+    mul_df = pd.DataFrame()
+
+    final_df = pd.DataFrame(columns=["testee_id", "file", "num", "testee_answer", "correct_answer"])
+
+    start = time.time()
+    categorize_qna(id_path)
+    end = time.time()
+
+    print("\ncategorize_qna eta: " + f"{end - start:.2f} sec")
+
+    mul_df = detect_multiple(id_path)
+    mul_df.sort_values(by=["num"], inplace=True)
+    print()
+    print_full(mul_df)
+
+    df = mul_df
+    df.sort_values(by=["num"], inplace=True)
+    print()
+    print_full(df)
+
+    final_df = dfToFinalDf(df)
+    print()
+    print_full(df)
+
+    final_df.to_excel(excel_writer=id_path+"/df.xlsx")
+
+    print_outro()
+
+    json_data = final_df.to_json(orient="records")
+    return json_data, 200
+
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=8080)

id.py

@@ -0,0 +1,100 @@
+import os
+import sys
+import easyocr
+
+import numpy as np
+
+from PIL import Image
+
+from utils import getNumText
+from utils import *
+
+sys.path.append(os.path.dirname(os.getcwd() + "/models/tamil_ocr/ocr_tamil"))
+from ocr_tamil.ocr import OCR
+
+
+### 텍스트 부분 잘라내기 함수 ###
+
+# 각 jpg에 적힌 코드 인식해서 이름 매칭
+# testee jpg df 생성
+# - testee_jpg_df = pd.DataFrame(columns=["file", "testee_id", "page"])
+# - name은 page가 1일 때만 인식 (학생 이름은 각 시험지 첫 장에만 적혀 있기 때문임)
+# 식별코드: testee_id - page (ex. 3-2라면, testee_id=3, page=2)
+
+### 오른쪽 상단 testee_name 인식 함수 ###
+def readTesteeName(img, reader):
+  x1, y1, x2, y2 = (610, 30, 750, 90)
+  cropped_img = img.crop((x1, y1, x2, y2))
+  image_np = np.array(cropped_img)
+
+  # tamilocr 사용
+  text = ""
+  ocr_text = OCR().predict(image_np)
+  text = getNumText(ocr_text)
+
+  return text
+
+
+### 텍스트에서 testee_id와 page를 추출하는 함수 ###
+def extractTesteeId(text):
+    if text:
+        text_split = text[0].split('-')
+        testee_id = text_split[0].strip()
+        page = text_split[1].strip() if len(text_split) > 1 else None
+        return testee_id, page
+    else:
+        return "", ""
+
+
+### testee_jpg_df에 id_match 추가
+def testeeIdJpgDf(df, testee_jpg_df, id_match):
+    # df = pd.DataFrame(columns=["index_id", "testee_id", "testee_name", "file", "page"])
+    for testee_jpg_df_idx, testee_jpg_df_row in testee_jpg_df.iterrows():
+        index_id = testee_jpg_df_row["index_id"]
+        testee_id = testee_jpg_df_row["testee_id"]
+        testee_name = id_match.loc[index_id, "testee_name"]
+        file = testee_jpg_df_row["file"]
+        page = testee_jpg_df_row["page"]
+        df.loc[len(df)] = [index_id, testee_id, testee_name, file, page]
+
+    return df  
+
+
+### 텍스트 부분 잘라내기 함수 (메인) ###
+def testeeCodeRecognition(jpg_file_path_list, testee_jpg_df):
+    # easyOCR 사용
+    reader = easyocr.Reader(['ko', 'en'])
+
+    # id_match 딕셔너리 초기화
+    id_match = pd.DataFrame(columns=["testee_id", "testee_name"])
+
+    # index_id
+    index_id = 0
+
+    # 폴더 내의 모든 파일에 대해 반복
+    for file in jpg_file_path_list:
+        # 이미지 파일 열기
+        img = Image.open(file)
+        img = img.resize((794,1123), Image.LANCZOS) # 인식 위치를 같게 만들기 위한 이미지 규격화.
+
+        # 왼쪽 상단 num_id와 page 인식
+        x1, y1, x2, y2 = (35, 35, 160, 90)
+        cropped_img = img.crop((x1, y1, x2, y2))
+        image_np = np.array(cropped_img)
+
+        # easyOCR 사용
+        text = reader.readtext(image_np, detail=0)
+        testee_id, page = extractTesteeId(text)
+
+        #오른쪽 상단 testee_name 인식
+        # page가 1인 경우 testee_id와 testee_name를 id_match에 딕셔너리로 추가
+        if page == "1":
+            testee_name = readTesteeName(img, reader)
+            id_match.loc[len(id_match)] = [testee_id, testee_name]
+            index_id += 1
+
+        testee_jpg_df.loc[len(testee_jpg_df)] = [index_id, file, testee_id, page]
+
+    id_match.index += 1
+
+    return testee_jpg_df, id_match