test_data.py

#Import Libraries
print("Importing libraries...")
import os, os.path
import numpy as np
import torch
import torch.nn as nn 
import torchvision
from torchvision import datasets, models, transforms
from sklearn.svm import SVC 
from sklearn.svm import LinearSVC
from sklearn.metrics import confusion_matrix
from PIL import Image
import pickle
print("Libraries imported")

#Directories
HOME_DIR = os.getcwd()
TESTDATA_DIR = os.path.join(HOME_DIR, 'Data/Test')

#Initialize arrays for testing the SVM
test_x = np.zeros((1,4096))
test_y = np.zeros((1), dtype='int64')

#Similar transforms as ImageNet on which model has been trained
from torchvision import transforms
transform = transforms.Compose([            #[1]
 transforms.Resize(224),                    #[2]
 transforms.ToTensor(),                     #[3]
 transforms.Normalize(                      #[4]
 mean=[0.485, 0.456, 0.406],                #[5]
 std=[0.229, 0.224, 0.225]                  #[6]
 )])


#Pretrained VGG19 without final softmax layer
model_conv = torchvision.models.vgg19(pretrained=True)
for param in model_conv.features.parameters():
    param.requires_grad = False
for param in model_conv.classifier.parameters():
    param.requires_grad = False
new_classifier = nn.Sequential(*list(model_conv.classifier.children())[:-1])
model_conv.classifier = new_classifier
model_conv.eval()

#Load the classifier
with open('linearSVM.pkl', 'rb') as fid:
    svm_model_linear = pickle.load(fid)

labels = ['neg', 'pos']
m = (len(os.listdir(os.path.join(TESTDATA_DIR, labels[0]))) 
    + len(os.listdir(os.path.join(TESTDATA_DIR, labels[1])))) #Size of testing data

j = 0 #Counter for printing progress
for label in labels:
    for img_name in os.listdir(os.path.join(TESTDATA_DIR, label)):
        print("Progress - ", int(j*100/m), "%", end ="\r")
        j += 1
        image = Image.open(os.path.join(TESTDATA_DIR, label, img_name))
        image_t = transform(image)
        batch_t = torch.unsqueeze(image_t, 0)
        out = model_conv(batch_t)
        np_out = np.asarray(out)
        test_x = np.concatenate((test_x,np_out))
        if label == 'neg':
            test_y = np.concatenate((test_y, np.array([0])))
        else:
            test_y = np.concatenate((test_y, np.array([1])))

#Remove the initial row of zeros present due to initialization    
test_x = test_x[1:,:]   
test_y = test_y[1:]    

print("----------Test Metrics----------")
print("Accuracy =", svm_model_linear.score(test_x, test_y)) 
svm_predictions = svm_model_linear.predict(test_x)
test_cm = confusion_matrix(test_y, svm_predictions) 
print("Confusion Matrix =\n", test_cm)