image color process

AI/app/controllers/example_controllers.py

@@ -1,6 +1,6 @@
 from flask import Blueprint, request, jsonify, Response
 from app.services import auto_draw
-import json, tensorflow as tf
+import json
 
 bp = Blueprint(name='example',
                import_name=__name__,
@@ -12,7 +12,7 @@
 
 
 @ai.route('/base64', methods=['POST'])
-def draw() -> str:
+def draw() -> Response:
     return jsonify(auto_draw.AIbyBase64(json.loads(request.get_json())))
 
 

AI/app/services/LoadImage.py → AI/app/services/ImageLoading.py

@@ -1,5 +1,4 @@
 import io, base64
-import os
 from PIL import Image
 import boto3
 import requests
@@ -18,8 +17,7 @@ def downloadFromS3(bucket, key):
     client = boto3.client('s3',
                           aws_access_key_id=AWS_ACCESS_KEY_ID,
                           aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
-                          region_name=AWS_DEFAULT_REGION
-                          )
+                          region_name=AWS_DEFAULT_REGION)
 
     file_name = 'downLoad.png'  # 다운될 이미지 이름
     # bucket = 'mufi-photo'  # 버켓 주소

AI/app/services/ImagePreprocessing.py

@@ -0,0 +1,30 @@
+from PIL import Image
+import numpy as np
+
+# 이미지를 흑백으로 변환
+def convertMono(img: Image):
+    fn = lambda x: 255 if x > 0 else 0
+    img = img.convert('L').point(fn, mode='1')
+    return img
+
+
+# 이미지 크기를 128*128로 변경 후 이미지를 배열화
+def resizing(img):
+    img = img.resize((128, 128))
+    img = img.convert('RGB')
+    return img
+
+
+# 이미지를 배열화
+def imageToArray(img):
+    img = np.array(img)
+    img = np.expand_dims(img, axis=0)
+    return img
+
+
+# 이미지를 AI에 맞게 전환
+def imageProcessing(img: Image):
+    img = convertMono(img)
+    img = resizing(img)
+    img = imageToArray(img)
+    return img

AI/app/services/auto_draw.py

@@ -1,29 +1,20 @@
 from keras.models import load_model
-import numpy as np
 import operator, os
-from app.services.LoadImage import decodeFromJsonToImage, downloadFromS3, downloadFromURL
+from app.services.ImageLoading import decodeFromJsonToImage, downloadFromS3, downloadFromURL
+from app.services.ImagePreprocessing import imageProcessing
 
 tags = 'airplane,apartment,apple,arm,arrow,axe,bag,baseball,basketball,bath,bed,bench,book,bottle,box,bread,broom,brush,bucket,bulb,cake,calendar,candle,candy,castle,chair,cherry,clock,cloud,coffee,coin,cookie,cup,cylinder,dice,dress,ear,earth,envelope,eraser,eye,finger,flag,flame,flask,flower,gamecontroller,ghost,gift,grape,hammer,heart,hospital,house,industry,injection,ladder,lake,leaf,leg,lightning,meat,megaphone,mic,money,monitor,mouse,mushroom,nail,nose,officebuilding,pants,peanut,pencil,police,pumpkin,rain,ribbon,rocket,ruler,school,shield,shirt,skirt,soccer,speaker,sprout,star,stethoscope,sun,swim,sword,television,tennis,tree,truck,umbrella,vehicle,volleyball,watch,wheel,windmill,zoom'.split(
     ',')
 
 
-# 이미지 크기를 128*128로 변경
-def resizing(image):
-    image = image.resize((128, 128))
-    image = image.convert('RGB')
-    image = np.array(image)
-    image = np.expand_dims(image, axis=0)
-    return image
-
-
-# AI를 이용하여 유사한 이미지 출력
+# AI를 이용하여 유사하다고 예상되는 이미지(이름) 출력
 def getPredict(img):
     saved_model = load_model(os.getcwd() + "/app/services/cnn1.h5")
     pre = saved_model.predict(img)
     return pre
 
 
-# AI 결과를 정확도에 따라 오름차순으로 정렬
+# AI 결과를 정확도에 따라 오름차 순으로 정렬
 def resultByDesc(result):
     x = {}
     for i in range(len(result)):
@@ -34,15 +25,16 @@ def resultByDesc(result):
 def getImageNameList(result):
     names = []
     for k, v in result:
-        print(k)
+        print(k,tags[k])
         names.append(tags[k])
     return names
 
+
 def AI_process(image):
-    resized = resizing(image)
-    pre_result = getPredict(resized)
-    result2 = resultByDesc(pre_result[0])
-    return getImageNameList(result2)
+    imgData = imageProcessing(image)
+    predict = getPredict(imgData)
+    result = resultByDesc(predict[0])
+    return getImageNameList(result)
 
 
 def AIbyBase64(json_data):