def plot_inkml(data, plot=True, save=False, classes=False):
filename = None
with open('colors','r') as c: # List of colors for each class, read from the 'colors' file
colors = eval(c.read())
if isinstance(data, str):
filename = data
data = parse_inkml(data) # Parses the inkml file if the input data is a filename (string) instead of some already processed inkml data (list)
for i in range(0, len(data['trace'])): # data['trace'] is a list of traces; each element contains a trace, and each trace has a list of [x, y] coordinates
x = []
y = []
for j in range(0, len(data['trace'][i])): # Returns j-th point of each trace
x.append(data['trace'][i][j][0]) # Returns x coordinate of the j-th point of each trace
y.append(data['trace'][i][j][1]) # Returns y coordinate of the j-th point of each trace
if classes:
plt.plot(x, y, color=colors[data['class'][i]], linewidth=0.6) # Plots current trace highlighting the trace's class with a specific color
else:
plt.plot(x, y, color='black', linewidth=0.6) # Plots current trace in plain black
plt.axis('equal') # Constrain proportions
plt.axis('off') # Remove axes from figure
if plot: # Only shows plot if plot = True
plt.show()
if save and filename is not None: # If save = True and the input data is a filename (not some already processed data), it saves the plot as a 640x480 px png image in a folder with the same name as the original file folder, with '_png' added at the end
plt.savefig(filename.partition('/')[0] + '_png/' + filename.partition('/')[2].replace('inkml', 'png'))
plt.cla() # Clears plot (otherwise subsequent plots would be drawn over the older ones)
def remove_random_elements_inkml(inkml_file, percentage=30):
data = parse_inkml(inkml_file) # Parses inkml file
perc = len(
data['group_id'].unique()
) * percentage / 100 # dati quanti group_id ho, faccio il 30%: es ho 25 gropu_id distinti e faccio il 30% di 25.
numRows = math.ceil(perc) # arrotondo la percentuale
random_number = [
] #lista vuota che servirà per segnare quali sono i numeri random usciti
for i in range(numRows):
# randrange gives you an integral value
irand = randrange(0, len(data['group_id'].unique(
))) #prendo un numero random tra 0 e il numero di group_id distinti
while irand in random_number:
irand = randrange(0, len(data['group_id'].unique()))
random_number.append(irand)
# Get names of indexes for which column group_id has value irand
indexNames = data[data['group_id'] == irand].index
# Delete these row indexes from dataFrame
data.drop(indexNames, inplace=True)
return data
def plot_inkml_remove_text(data, save=True):
print(data)
filename = None
if not ('b' in data): # controlla che il file su cui fare i bb sia un file di train e NON di test
with open('colors','r') as c: # List of colors for each class, read from the 'colors' file
colors = eval(c.read())
if isinstance(data, str):
filename = data
data = parse_inkml(data) # Parses the inkml file if the input data is a filename (string) instead of some already processed inkml data (list)
for i in range(0, len(data['trace'])): # data['trace'] is a list of traces; each element contains a trace, and each trace has a list of [x, y] coordinates
x = []
y = []
for j in range(0, len(data['trace'][i])): # Returns j-th point of each trace
x.append(data['trace'][i][j][0]) # Returns x coordinate of the j-th point of each trace
y.append(data['trace'][i][j][1]) # Returns y coordinate of the j-th point of each trace
if not(data['class'][i] == 'Text'): # se ha come classe Text non viene plot
plt.plot(x, y, color='black', linewidth=0.6)
plt.axis('equal') # Constrain proportions
plt.axis('off') # Remove axes from figure
if save and filename is not None: # If save = True and the input data is a filename (not some already processed data), it saves the plot as a 640x480 px png image in a folder with the same name as the original file folder, with '_png' added at the end
fig1 = plt.gcf()
#plt.show()
plt.draw()
#plt.cla()
fig1.savefig('path/FCinkML_aug/' + filename.partition('/')[2].replace('FCinkML', 'FCinkML_aug').replace('writer', 'writer_augm_').replace('inkml', 'png')) #cambio il nome in modo da non sovrascrivere le immagini
#fig1.savefig(filename.partition('/')[0] + '_aug/' + filename.partition('/')[2].replace('inkml', 'png'))
plt.cla() # Clears plot (otherwise subsequent plots would be drawn over the older ones)
else:
print('Test File')
# Example svg to inkml conversion and inkml plotting
n = 1
for i in range(0, n + 1):
svg_file = './NeoSmartpenM1_demo/neo_smtpen_m1_demo_' + str(
i) + '.svg' # svg file to be converted to inkml
inkml_file = './NeoSmartpenM1_demo/neo_smtpen_m1_demo_' + str(
i) + '.inkml' # Converted file name
svg2inkml(svg_file, inkml_file) # Converts svg file to inkml
plot_inkml(inkml_file) # Plots the returned data
# Other test plots
show = True
plot_inkml(
parse_inkml('./FCinkML/writer3_5.inkml'),
plot=show) # Plots a random inkml file from the dataset (by parsed data)
plot_inkml(
'./FCinkML/writer3_5.inkml',
plot=show) # Plots a random inkml file from the dataset (by filename)
plot_inkml(
parse_inkml('./FCinkML/writer3_5.inkml'), plot=show, classes=True
) # Plots a random inkml file from the dataset (by parsed data) and highlights classes
plot_inkml(
'./FCinkML/writer3_5.inkml', plot=show, classes=True
) # Plots a random inkml file from the dataset (by filename) and highlights classes
from draw_bb import draw_bb
draw_bb('./FCinkML/writer3_5.inkml', show=True)
def bounding_boxes(inkml_file, save=False, plot=False):
annotation_file = 'noText_annotation.csv' # Edit this line with your own annotation file
data = parse_inkml(inkml_file) # Parses inkml file
cols = ['x_min', 'x_max', 'y_min', 'y_max', 'class'] # Columns of bboxs dataframe
bboxs = pd.DataFrame(columns=cols) # Dataframe containing all coordinates of all bounding boxes, plus the class for each bounding box
#------------------------------------------------------------------------
if plot: # If plot = True, plot the original figure first
with open('colors','r') as c: # List of colors for each class, read from the 'colors' file
colors = eval(c.read())
for i in range(0, len(data['trace'])): # data['trace'] is a list of traces; each element contains a trace, and each trace has a list of [x, y] coordinates
x = [] # Contains current trace x coordinates
y = [] # Contains current trace y coordinates
for j in range(0, len(data['trace'][i])): # Return j-th point of each trace
x.append(data['trace'][i][j][0]) # Returns x coordinate of the j-th point of each trace
y.append(data['trace'][i][j][1]) # Returns y coordinate of the j-th point of each trace
#plt.plot(x, y, color='black', linewidth=0.6) # Plot current trace in plain black
if not (data['class'][i] == 'Text'): # se ha come classe Text non viene plot
plt.plot(x, y, color='black', linewidth=0.6)
plt.axis('equal') # Constrain proportions
plt.axis('off') # Remove axes from figure
# ------------------------------------------------------------------------
# Crea bb per ogni figura
# Escludo le tracce che appartengono alla classe text
for k in range(0, len(data['group_id'].drop_duplicates())): # For each unique trace group of the inkml file...
group_x = [] # Contains all x coordinates of all traces in a specific trace group
group_y = [] # Contains all y coordinates of all traces in a specific trace group
for i in range(0, len(data['trace'])): # ...for each trace in the dataset...
for j in range(0, len(data['trace'][i])): # ...for each pair of [x, y] coordinates in trace i...
if data['group_id'][i] == k: # ...get the [x, y] coordinates in two separate arrays
if not (data['class'][i] == 'Text'):
label = data['class'][i] # Identify the class of the current bounding box
group_x.append(data['trace'][i][j][0]) # Add all x coordinates of trace i to the group x coordinates
group_y.append(data['trace'][i][j][1]) # Add all y coordinates of trace i to the group y coordinates
group_x.sort() # Sort the x coordinates array
group_y.sort() # Sort the y coordinates array
#----------------------------------
bboxs = bboxs.append({'x_min': group_x[0], 'x_max': group_x[len(group_x) - 1], 'y_min': group_y[0], 'y_max': group_y[len(group_y) - 1], 'class': label}, ignore_index=True) # Save minimum and maximum [x, y] coordinates to the bboxs dataframe
# Add +1 to the maximum coordinates and -1 to the minimum to get
x_coord = [group_x[0], group_x[len(group_x)-1], group_x[len(group_x)-1], group_x[0], group_x[0]] # Contain all x coordinates of a bounding box, plus the first one repeated to be able to draw a rectangle over the figure
y_coord = [group_y[len(group_y)-1], group_y[len(group_y)-1], group_y[0], group_y[0], group_y[len(group_y)-1]] # Contain all x coordinates of a bounding box, plus the first one repeated to be able to draw a rectangle over the figure
if plot: # If plot = True, plot one by one the resulting bounding boxes over the original figure from data coordinates (not pixels!), that has also been plotted if plot = True
plt.plot(x_coord, y_coord, color=colors[label], linewidth=1)
pboxs = pd.DataFrame(columns=cols) # Create a new dataframe containing the pixel coordinates of the bounding boxes
# print('Calculating pixel coordinates for ' + filename, end='')
for i in range(0, len(bboxs)): # Transform data coordinates in bboxs to pixel coordinates
pboxs = transform_coord(data, bboxs, i, pboxs)
# print(' done.')
if plot: # If plot = True, show the generated plot; the whole plot thing is for debug purposes only (and nice, coloured images)
plt.show()
pboxs['class'] = bboxs[['class']].copy() # Copy the labels from the original bboxs dataframe, as they will not change later
if save: # If save = True, append the bounding box pixel coordinates to the specified annotation file
#DA METTERE .replace('writer', 'writer_augm_')
filename = inkml_file.split('/')[len(inkml_file.split('/'))-1].replace('writer', 'writer_augm_').replace('.inkml', '.png') # Get the image file name from the inkml file -------
pboxs.insert(0, 'filename', 'path/' + filename) # Insert a new column to pboxs in position 0, containing the processed file name
cols = ['filename', 'x_min', 'y_min', 'x_max', 'y_max', 'class'] # csv file columns
pboxs.to_csv(annotation_file, columns=cols, header=False, index=False, mode='a') # Add data to the specified annotation file
print('Annotation data for ' + inkml_file.split('/')[len(inkml_file.split('/'))-1] + ' saved to ' + annotation_file + '.')
return pboxs, bboxs
from parse_inkml import parse_inkml
from transform_coord import transform_coord
parser = argparse.ArgumentParser(description='Returns bounding boxes pixel coordinates for the specified file.') # Parser definition
parser.add_argument('-file', action='store', dest='file') # Declare an argument called '-inkml_file' that is going to be stored in the 'inkml_file' variable
parser.add_argument('-annotation', action='store', dest='annotation') # Declare an argument called '-annotation' that is going to be stored in the 'annotation' variable
inkml_file = parser.parse_args().file # Parse the command line arguments and store the value in the 'inkml_file' variable
annotation_file = parser.parse_args().annotation # Parse the command line arguments and store the value in the 'annotation_file' variable
filename = inkml_file.split('/')[len(inkml_file.split('/'))-1].replace('.inkml', '.png') # Get the image file name from the inkml file
print('Working on ' + filename + '... ', end='')
data = parse_inkml(inkml_file) # Parses inkml file
cols = ['x_min', 'x_max', 'y_min', 'y_max', 'class'] # Columns of bboxs dataframe
bboxs = pd.DataFrame(columns=cols) # Dataframe containing all coordinates of all bounding boxes, plus the class for each bounding box
for k in range(0, len(data['group_id'].drop_duplicates())): # For each unique trace group of the inkml file...
group_x = [] # Contains all x coordinates of all traces in a specific trace group
group_y = [] # Contains all y coordinates of all traces in a specific trace group
for i in range(0, len(data['trace'])): # ...for each trace in the dataset...
for j in range(0, len(data['trace'][i])): # ...for each pair of [x, y] coordinates in trace i...
if data['group_id'][i] == k: # ...get the [x, y] coordinates in two separate arrays
label = data['class'][i] # Identify the class of the current bounding box
group_x.append(data['trace'][i][j][0]) # Add all x coordinates of trace i to the group x coordinates
group_y.append(data['trace'][i][j][1]) # Add all y coordinates of trace i to the group y coordinates