Upload 3 files

Logs/detectnameanddistance.py

@@ -0,0 +1,51 @@
+import cv2
+import numpy as np
+def render(Session,input_name,output_name,FacesEmbedding,results,FaceImage,persons):
+    for res in results.face_landmarks:
+        
+        x_=int(res[145].x*FaceImage.shape[1])
+        y_=int(res[145].y*FaceImage.shape[0])
+        x2_=int(res[374].x*FaceImage.shape[1])
+        y2_=int(res[374].y*FaceImage.shape[0])
+        w=np.sqrt((x_-x2_)**2+(y_-y2_)**2)
+        W=6.3
+        f = 840
+        d = (W * f) / w
+        x=int(res[356].x*FaceImage.shape[1])
+        y=int(res[152].y*FaceImage.shape[0])
+        x2=int(res[162].x*FaceImage.shape[1])
+        y2=int(res[338].y*FaceImage.shape[0])
+        if x<FaceImage.shape[1]-10:
+            x+=10
+        if y>FaceImage.shape[0]-10:
+            y+=10
+        if x2>10:
+            x2-=10
+        if y2>10:
+            y2-=10
+        
+        
+        modelimg=FaceImage[y2:y,x2:x]
+        
+        
+        if modelimg.size<9:
+            continue
+        modelimg=cv2.resize(modelimg,(224,224)).astype(np.float32)
+        modelimg=modelimg/255
+        
+        distances=[]
+        if d>0:
+            for index,name in enumerate(persons):
+                output=np.squeeze(Session.run([output_name],{f"{input_name}":np.expand_dims(modelimg,axis=0).astype(np.float16)})[0])
+                personimpeding=FacesEmbedding[name].values
+                distance=np.sum(np.power(output-personimpeding,2))
+                distances.append(distance)
+            name=persons[np.argmin(distances)]
+            distance=distances[np.argmin(distances)]
+            if distance <0.3:
+                return x_,y_,x2_,y2,name,d
+                
+            else:
+                return x_,y_,x2_,y2,"UnKnow",d
+                
+

main.py

@@ -0,0 +1,53 @@
+from fastapi import FastAPI,UploadFile
+from PIL import Image
+import cv2
+import mediapipe as mp
+import numpy as np
+import pandas as pd
+from io import BytesIO
+import onnxruntime as ort
+from Logs.detectnameanddistance import render
+
+
+
+FacesEmbedding=pd.read_csv("./Models/FacesMeanEmbeddings.csv",index_col=0)
+persons=list(FacesEmbedding.columns)
+model_path="./Models/FaceModelV5.onnx"
+EP_list = [ 'CPUExecutionProvider']
+Session = ort.InferenceSession(model_path,providers=EP_list)
+input_name = Session.get_inputs()[0].name
+output_name=Session.get_outputs()[0].name
+MediapipeModelPath="./Models/face_landmarker.task"
+BaseOptions=mp.tasks.BaseOptions
+FaceLandMarker=mp.tasks.vision.FaceLandmarker
+FaceLandMarkerOptions=mp.tasks.vision.FaceLandmarkerOptions
+VisionRunningMode=mp.tasks.vision.RunningMode
+FaceLandMarkerResult=mp.tasks.vision.FaceLandmarkerResult
+options=FaceLandMarkerOptions(base_options=BaseOptions(model_asset_path=MediapipeModelPath),running_mode=VisionRunningMode.IMAGE)
+landmarker= FaceLandMarker.create_from_options(options)
+
+App=FastAPI()
+
+@App.post("/upload")
+async def detect(img:UploadFile):
+    
+    image=np.array(Image.open(BytesIO(img.file.read())))
+    mp_img=mp.Image(image_format=mp.ImageFormat.SRGB,data=image)
+    result=landmarker.detect(mp_img)
+    
+    if len(result.face_landmarks)==0:
+        return {"state":False,"message":"No Face Found","distance":0,"name":"null"}
+    x1,y1,x2,y2,name,distance=render(Session,input_name,output_name,FacesEmbedding,result,mp_img.numpy_view(),persons)
+    return {"state":True,"message":"null","distance":distance,"name":name,"x1":x1,"x2":x2,"y1":y1,"y2":y2}
+
+# def getHighScore(Scores):
+    
+#     if type(Scores)==dict:
+#         return False
+#     scorep=0.00
+#     label="0-10"
+#     for Score_ in Scores:
+#         if Score_["score"] > scorep:
+#             scorep=float(Score_["score"])
+#             label=Score_["label"]
+#     return label

requirements.txt

@@ -0,0 +1,10 @@
+fastapi
+mediapipe
+numpy
+opencv_contrib_python
+opencv_python
+pandas
+uvicorn
+Requests
+onnxruntime
+python-multipart