def main():
# Get input arguments
wddOutputRoot = sys.argv[1]
inputModelFile = sys.argv[2]
resultFolder = sys.argv[3]
print('Folder with unfiltered dances is "', wddOutputRoot)
print('Input model file is "', inputModelFile)
kM = kerasModel.KerasModel()
model = kM.getModel()
model.load_weights(inputModelFile)
# Init Confusion Matrix
CM = np.zeros((2,2))
progress = 0
# Traverse folder structure and build matrix for every single dance so that CNN can test it
# Set the directory you want to start from
rootDir = wddOutputRoot
for dirName, subdirList, fileList in os.walk(rootDir):
print(dirName)
if 'orient.png' in fileList:
image_list = []
for fname in glob.glob(dirName + '/image*.png'):
if fname != 'orient.png':
im = scipy.misc.imread(fname)[:, :, 1]
image_list.append(im)
image_array = np.asarray(image_list)
pred = classify_dance(image_array, model, kM.get_image_count())
CM = update_confusion_matrix(pred, CM, dirName, resultFolder)
progress += 1
print(progress)
print(CM)
def main():
validationFolder = sys.argv[1]
inputModelFile = sys.argv[2]
outputFolder = sys.argv[3]
print('Validation folder is "', validationFolder)
print('Input model file is "', inputModelFile)
print('Output folder is "', outputFolder)
kM = kerasModel.KerasModel()
model = kM.getModel()
model.load_weights(inputModelFile)
# Init Confusion Matrix
CM = np.zeros((2, 2))
progress = 0
# traverse folder structure and build matrix for every single dance so that CNN can test it
# Set the directory you want to start from
rootDir = validationFolder
for dirName, subdirList, fileList in os.walk(rootDir):
print("Found directory: %s" % dirName)
if "gt.csv" in fileList:
with open(dirName + "/gt.csv", "rt") as csvfile:
spamReader = csv.reader(csvfile, delimiter=" ", quotechar="|")
Y = 0
for row in spamReader:
Y = row[0]
if Y == "j":
Y = 1
elif Y == "n":
Y = 0
else:
Y = -1
continue
image_list = []
for fname in glob.glob(dirName + "/image*.png"):
im = scipy.misc.imread(fname)[:, :, 1]
image_list.append(im)
image_array = np.asarray(image_list)
pred = classify_dance(image_array, model, kM.get_image_count())
CM = update_confusion_matrix(pred, CM, Y, dirName, outputFolder)
progress += 1
print(progress)
print(CM)
def main():
# Get input arguments
validationFolderRoot = sys.argv[1]
inputModelFile = sys.argv[2]
print('Validation folder root is "', validationFolderRoot)
print('Input model file is "', inputModelFile)
kM = kerasModel.KerasModel()
model = kM.getModel()
model.load_weights(inputModelFile)
# Init Confusion Matrix
CM = np.zeros((2,2))
progress = 0
# Traverse folder structure and build matrix for every single dance so that CNN can test it
# Set the directory you want to start from
rootDir = validationFolderRoot
for dirName, subdirList, fileList in os.walk(rootDir):
print('Found directory: %s' % dirName)
if 'gt.csv' in fileList:
with open(dirName+'/gt.csv', 'rt') as csvfile:
spamReader = csv.reader(csvfile, delimiter=' ', quotechar='|')
Y = 0
for row in spamReader:
Y = row[0]
if Y == 'j':
Y = 1
elif Y == 'n':
Y = 0
else:
Y = -1
continue
image_list = []
for fname in glob.glob(dirName + '/image*.png'):
im = scipy.misc.imread(fname)[:, :, 1]
image_list.append(im)
image_array = np.asarray(image_list)
pred = classify_dance(image_array, model, kM.get_image_count())
CM = update_confusion_matrix(pred, CM, Y)
progress += 1;
print(progress)
print(CM)