kernel = Pyramid_Kernel(k=512, pyramid_levels=pyramid_levels)
mySVM = SVM(kernel=kernel.pyramid_kernel)
else:
# hardcode, load best parameters found with cross validation
with open(
"/imatge/mgorriz/work/master/models/session02/test1/best_params_svm.pkl",
'rb') as file:
best_params_svm = pickle.load(file)
mySVM = SVM(kernel=best_params_svm['kernel'],
C=best_params_svm['C'],
gamma=best_params_svm['gamma'])
model = mySVM.load_model(args.model_path)
test_data = InputData.get_test_data()
if args.do_compute_features:
#descriptors, descriptors_idx = sift_descriptors.extract_features_simple(data_dictionary=test_data)
test_descriptors_path = "/imatge/mgorriz/work/master/models/session02/test2/test_descriptors.npy"
descriptors, descriptors_idx = \
sift_descriptors.extract_features_simple(data_dictionary=test_data,
load=True, path=test_descriptors_path)
codebook = bow_descriptor.load(
os.path.join(args.codebook_path, 'codebook.pkl'), 'codebook')
test_visual_words = bow_descriptor.extract_features(
descriptors, codebook, descriptors_idx)
bow_descriptor.save(
display=True)
# save model
myKNN.save_model(model, args.model_path)
else:
print("Invalid train method")
print('Training finished: done in ' + str(time.time() - start_time) +
' secs')
elif args.do_test:
start_time = time.time()
test_data = InputData.method_data_dictionary(InputData.get_test_data(),
'test')
model = myKNN.load_model(args.model_path)
# test model
predictions = myKNN.predict(test_data['filenames'],
model,
display=True)
myEvaluation.accuracy(test_data['labels'], predictions, display=True)
myEvaluation.confusion_matrix(test_data['labels'],
predictions,
display=True)
print('Test finished: done in ' + str(time.time() - start_time) +
' secs')
