predictor

predictor.py

@@ -0,0 +1,57 @@
+import os
+import sys
+import pandas as pd
+import numpy as np
+from skimage.io import imread, imshow, imsave
+from skimage.transform import resize
+from sklearn.preprocessing import StandardScaler
+from sklearn.decomposition import PCA
+from sklearn.model_selection import train_test_split
+from sklearn import svm
+from sklearn import metrics
+import pickle
+from skimage import img_as_ubyte
+
+tempdir = "predictor_temp"
+
+if len(sys.argv) < 3:
+    print("no input image")
+    exit(0)
+
+if len(sys.argv) < 2:
+    print("no input model")
+    exit(0)
+
+svm_model = pickle.load(open(sys.argv[1], 'rb'))
+
+os.system('mkdir -p ' + tempdir)
+os.system("python3 face_cutter.py " + sys.argv[2] + " " + tempdir )
+img_files = [name for name in os.listdir(tempdir) if  not os.path.isdir(os.path.join(tempdir, name)) ]
+print(img_files)
+cutf = tempdir +"/"+img_files[0]
+os.system("python3 hogger.py " + cutf + " " + tempdir )
+os.system("rm " + cutf)
+img_files = [name for name in os.listdir(tempdir) if  not os.path.isdir(os.path.join(tempdir, name)) ]
+hogf = tempdir + "/" + img_files[0]
+
+imgdat = imread(hogf, as_gray=True)
+os.system("rm -rf " + tempdir)
+imgdat = resize(imgdat, (64,64))
+
+imgdat = img_as_ubyte(imgdat)
+
+
+flat_imgdat = np.array( imgdat ).flatten()
+print(flat_imgdat.shape)
+X = np.array(flat_imgdat)
+X = X.reshape(-1,1)
+X = StandardScaler().fit_transform(X)
+print(X.shape,X)
+# pca = PCA(n_components=128)
+# pcaofX = pca.fit_transform(X)
+
+
+res = svm_model.predict(X.T)
+
+print("Prediction:",res)
+

trainer.py

@@ -6,6 +6,7 @@ from sklearn.decomposition import PCA
 from sklearn.model_selection import train_test_split
 from sklearn import svm
 from sklearn import metrics
+import pickle
 
 if len(sys.argv) < 2:
     print("no input csv file")
@@ -27,19 +28,24 @@ print(Y.shape)
 
 X = StandardScaler().fit_transform(X)
 print(X)
-pca = PCA(n_components=50)
+# pca = PCA(n_components=128)
-pcaofX = pca.fit_transform(X)
+# pcaofX = pca.fit_transform(X)
-print("shapeofX after pca",pcaofX.shape, ", cum Sum of variance ratio",pca.explained_variance_ratio_.cumsum()[-1])
+# print("shapeofX after pca",pcaofX.shape, ", cum Sum of variance ratio",pca.explained_variance_ratio_.cumsum()[-1])
 
-# pcaofX = X
+pcaofX = X
 
-X_train, X_test, Y_train, Y_test = train_test_split(pcaofX, Y, test_size=0.3,random_state=109)
+X_train, X_test, Y_train, Y_test = train_test_split(pcaofX, Y, test_size=0.1,random_state=109)
 
 print(X_train.shape)
 
-classifier = svm.SVC(kernel="linear")
+svm_model = svm.SVC(kernel="linear")
-classifier.fit(X_train,Y_train)
+svm_model.fit(X_train,Y_train)
 
-pred = classifier.predict(X_test)
+pred = svm_model.predict(X_test)
+print(pred)
+
+print("Accuracy:",metrics.accuracy_score(Y_test, pred))
+
+
+pickle.dump(svm_model, open("svm_model.sav", 'wb'))
 
-print("Accuracy:",metrics.accuracy_score(Y_test, pred))