def preprocess_datasets(self, path_dataset, groupStage):
PATH_DATA = os_path.join(path_dataset, "4")
print("(INFO) EVALUATING DATASET ...")
path_img = sorted(listdir(PATH_DATA))
if path_img == []:
return -1, -1
num_img = len(path_img)
# Histogram of all images in folder
hChannel = []
sChannel = []
vChannel = []
for image_path in path_img:
img = imread(os_path.join(PATH_DATA, image_path))
img = resize(img, (6000, 4000))
img = img[500:-500, 750:-750, :]
# HSV channel
img = cvtColor(img, COLOR_BGR2HSV)
# HSV histogram
h = calcHist([img], [0], None, [256], [0, 256]).reshape(256, )
s = calcHist([img], [1], None, [256], [0, 256]).reshape(256, )
v = calcHist([img], [2], None, [256], [0, 256]).reshape(256, )
hChannel.append(h)
sChannel.append(s)
vChannel.append(v)
# Compute dissimilarity
maxI = 0
for i in range(num_img):
one = []
for j in range(num_img):
c1 = np_sum(
np_absolute(hChannel[j] - hChannel[i])) / (HEIGHT * WIDTH)
c2 = np_sum(
np_absolute(sChannel[j] - sChannel[i])) / (HEIGHT * WIDTH)
c = (c1 + c2) / 2
if c > maxI:
maxI = c
save = [i, j]
img0 = path_img[save[0]]
img1 = path_img[save[1]]
imgSample1 = os_path.join(PATH_DATA, img0)
imgSample2 = os_path.join(PATH_DATA, img1)
return imgSample1, imgSample2
def predict_image(path_of_image, groupStage):
path_of_model = os_path.join("./CUSTOMIZE_4_USER/MODEL_TRAINING",
groupStage, groupStage + ".pth")
path_of_feature = os_path.join("./CUSTOMIZE_4_USER/MODEL_TRAINING",
groupStage, groupStage + ".npz")
start_time = time()
model = NeuralNet(input_size, hidden_size, num_classes).to(device)
model.load_state_dict(load(path_of_model))
data = np_load(path_of_feature)
[h_max, s_max, v_max] = data['data_max']
[h_min, s_min, v_min] = data['data_min']
img = imread(path_of_image)
img = resize(img, (6000, 4000))
img = img[500:-500, 750:-750, :]
img = cvtColor(img, COLOR_BGR2HSV)
hchan, schan, vchan = split(img)
h_hist = calcHist([img], [0], None, [256], [0, 256]).reshape(256, )
s_hist = calcHist([img], [1], None, [256], [0, 256]).reshape(256, )
v_hist = calcHist([img], [2], None, [256], [0, 256]).reshape(256, )
hMean = np_mean(hchan) / 255
DPV_h_max = np_sum(np_absolute(h_hist - h_max)) / (HEIGHT * WIDTH)
DPV_h_min = np_sum(np_absolute(h_hist - h_min)) / (HEIGHT * WIDTH)
sMean = np_mean(schan) / 255
DPV_s_max = np_sum(np_absolute(s_hist - s_max)) / (HEIGHT * WIDTH)
DPV_s_min = np_sum(np_absolute(s_hist - s_min)) / (HEIGHT * WIDTH)
vMean = np_mean(vchan) / 255
DPV_v_max = np_sum(np_absolute(v_hist - v_max)) / (HEIGHT * WIDTH)
DPV_v_min = np_sum(np_absolute(v_hist - v_min)) / (HEIGHT * WIDTH)
correlation = np_corrcoef(h_hist, s_hist)[0][1]
#image_feature = np_array((hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, vMean, DPV_v_max, DPV_v_min))
image_feature = np_array((hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max,
DPV_s_min, correlation))
image_feature = from_numpy(image_feature).to(device).float().view(
1, input_size)
with no_grad():
out_predict = model(image_feature)
_, predicted_result = torch_max(out_predict.data, 1)
original = Tensor([[1, 33, 66, 99]])
# Round xx.xx %
percentage_result = np_round(
mm(out_predict.view(1, num_classes), original.view(num_classes,
1)).item(), 2)
# Processed time
processedTime = np_round(time() - start_time, 2)
#print("Time ",processedTime)
return percentage_result, processedTime
def process_feature(list_path, labelFeature):
list_dir = sorted(listdir(list_path))
if list_dir == []:
return -1
for image_path in list_dir:
name_image = os_path.join(list_path, image_path)
if name_image == imgSample1 or name_image == imgSample2:
continue
img = imread(name_image)
img = resize(img, (6000, 4000))
img = img[500:-500, 750:-750, :]
img = cvtColor(img, COLOR_BGR2HSV)
hchan, schan, vchan = split(img)
h_hist = calcHist([img], [0], None, [256],
[0, 256]).reshape(256, )
s_hist = calcHist([img], [1], None, [256],
[0, 256]).reshape(256, )
v_hist = calcHist([img], [2], None, [256],
[0, 256]).reshape(256, )
hMean = np_mean(hchan) / 255
DPV_h_max = np_sum(
np_absolute(h_hist - h_max)) / (HEIGHT * WIDTH)
DPV_h_min = np_sum(
np_absolute(h_hist - h_min)) / (HEIGHT * WIDTH)
sMean = np_mean(schan) / 255
DPV_s_max = np_sum(
np_absolute(s_hist - s_max)) / (HEIGHT * WIDTH)
DPV_s_min = np_sum(
np_absolute(s_hist - s_min)) / (HEIGHT * WIDTH)
vMean = np_mean(vchan) / 255
DPV_v_max = np_sum(
np_absolute(v_hist - v_max)) / (HEIGHT * WIDTH)
DPV_v_min = np_sum(
np_absolute(v_hist - v_min)) / (HEIGHT * WIDTH)
correlation = np_corrcoef(h_hist, s_hist)[0][1]
# variable = [hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, vMean, DPV_v_max, DPV_v_min]
variable = [
hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min,
correlation
]
feature.append(variable)
labels.append([labelFeature])
def process_feature(list_path, labelFeature):
print("Extracting...")
list_dir = sorted(listdir(list_path))
if list_dir == []:
return -1
for image_path in list_dir:
name_image = os_path.join(list_path, image_path)
if name_image == imgSample1 or name_image == imgSample2:
continue
img = imread(name_image)
img = resize(img, (6000,4000))
img = img[500:-500, 750:-750, :]
img = cvtColor(img, COLOR_BGR2HSV)
hchan, schan, vchan = split(img)
h_hist = calcHist([img], [0], None, [256], [0,256]).reshape(256,)
s_hist = calcHist([img], [1], None, [256], [0,256]).reshape(256,)
v_hist = calcHist([img], [2], None, [256], [0,256]).reshape(256,)
# 7 feature consist of :
# + Compute mean value pixel of H channel
# + Dissilarity with H channel of "max" image
# + Dissilarity with H channel of "min" image
# + Compute mean value pixel of S channel
# + Dissilarity with S channel of "max" image
# + Dissilarity with S channel of "min" image
# + Correlation between histogram of H and S channel
hMean = np_mean(hchan)/255
DPV_h_max = np_sum(np_absolute(h_hist - h_max))/(HEIGHT*WIDTH)
DPV_h_min = np_sum(np_absolute(h_hist - h_min))/(HEIGHT*WIDTH)
sMean = np_mean(schan)/255
DPV_s_max = np_sum(np_absolute(s_hist - s_max))/(HEIGHT*WIDTH)
DPV_s_min = np_sum(np_absolute(s_hist - s_min))/(HEIGHT*WIDTH)
vMean = np_mean(vchan)/255
DPV_v_max = np_sum(np_absolute(v_hist - v_max))/(HEIGHT*WIDTH)
DPV_v_min = np_sum(np_absolute(v_hist - v_min))/(HEIGHT*WIDTH)
correlation = np_corrcoef(h_hist, s_hist)[0][1]
# variable = [hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, vMean, DPV_v_max, DPV_v_min]
variable = [hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, correlation]
feature.append(variable)
labels.append([labelFeature])
def AmpTrim(self):
return np_absolute(self.ampTrim)
def Amp(self):
return np_absolute(self.amp)
def predict_image(path_of_image, groupStage):
path_of_model = os_path.join("./CUSTOMIZE_4_USER/MODEL_TRAINING", groupStage, groupStage+".pth")
path_of_feature = os_path.join("./CUSTOMIZE_4_USER/MODEL_TRAINING", groupStage, groupStage+".npz")
hidden_path = os_path.join("./CUSTOMIZE_4_USER/MODEL_TRAINING", groupStage, groupStage+"_hidden.txt")
# if hidden number file is not existing, return 0
try:
with open(hidden_path,"r") as f:
hidden_size = int(f.readline())
except FileNotFoundError:
print("ERROR: No hidden number file in folder")
return 0.0,0
# Calculate processing time
start_time = time()
model = NeuralNet(input_size, hidden_size, num_classes).to(DEVICE)
model.load_state_dict(load(path_of_model))
data = np_load(path_of_feature)
[h_max, s_max, v_max] = data['data_max']
[h_min, s_min, v_min] = data['data_min']
img = imread(path_of_image)
img = resize(img, (6000,4000))
img = img[500:-500, 750:-750, :]
img = cvtColor(img, COLOR_BGR2HSV)
hchan, schan, vchan = split(img)
h_hist = calcHist([img], [0], None, [256], [0,256]).reshape(256,)
s_hist = calcHist([img], [1], None, [256], [0,256]).reshape(256,)
v_hist = calcHist([img], [2], None, [256], [0,256]).reshape(256,)
# 7 features consist of :
# + Compute mean value pixel of H channel
# + Dissilarity with H channel of "max" image
# + Dissilarity with H channel of "min" image
# + Compute mean value pixel of S channel
# + Dissilarity with S channel of "max" image
# + Dissilarity with S channel of "min" image
# + Correlation between histogram of H and S channel
hMean = np_mean(hchan)/255
DPV_h_max = np_sum(np_absolute(h_hist - h_max))/(HEIGHT*WIDTH)
DPV_h_min = np_sum(np_absolute(h_hist - h_min))/(HEIGHT*WIDTH)
sMean = np_mean(schan)/255
DPV_s_max = np_sum(np_absolute(s_hist - s_max))/(HEIGHT*WIDTH)
DPV_s_min = np_sum(np_absolute(s_hist - s_min))/(HEIGHT*WIDTH)
vMean = np_mean(vchan)/255
DPV_v_max = np_sum(np_absolute(v_hist - v_max))/(HEIGHT*WIDTH)
DPV_v_min = np_sum(np_absolute(v_hist - v_min))/(HEIGHT*WIDTH)
correlation = np_corrcoef(h_hist, s_hist)[0][1]
#image_feature = np_array((hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, vMean, DPV_v_max, DPV_v_min))
image_feature = np_array((hMean, DPV_h_max, DPV_h_min, sMean, DPV_s_max, DPV_s_min, correlation))
image_feature = from_numpy(image_feature).to(DEVICE).float().view(1, input_size)
with no_grad():
out_predict = model(image_feature)
# Round xx.xx %
percentage_result = np_round(out_predict.item()*99, 2)
if percentage_result >99.99:
percentage_result = 99.99
if percentage_result <1.0:
percentage_result = 1.0
# Processed time
processedTime = np_round(time()-start_time, 2)
return percentage_result, processedTime