In [106]:

    

import numpy as np
from svgpathtools import svg2paths, Path, Line, wsvg
import svgpathtools
from rdp import rdp
from svg.path import parse_path
import os
import pandas as pd


    

In [140]:

    

# define paths
data_dir = './svg'
save_dir = './spline_coords'
start_from = 'cat'


    

In [141]:

    

def read_svg_file(filename):
    paths, attributes =  svg2paths(filename)
    return paths


    

In [143]:

    

all_classes = sorted([i for i in os.listdir(data_dir) if os.path.isdir(os.path.join(data_dir,i))])
assert len(all_classes)==125

ind = np.where(np.array(all_classes)==start_from)[0][0]
all_classes = all_classes[ind:]

# loop through classes
for c in all_classes:
    class_path = os.path.join(path_to_svg,c)
    all_sketches = [i for i in os.listdir(class_path) if i.split('.')[1]=='svg']
    
    # initialize
    start_x = []
    start_y = []
    c1_x = []
    c1_y = []
    c2_x = []
    c2_y = []
    end_x = []
    end_y = []
    stroke_num = []
    spline_num = []
    stroke_counter = 0
    spline_counter = 0

    class_name = []
    photo_name = []
    sketch_name = []

    path_to_invalid = os.path.join(class_path,'invalid.txt')
    with open(path_to_invalid) as f:
        invalid = set([line[:-2] for line in f.readlines() if line[0]=='n'])
    cat = class_path.split('/')[-1]
    # loop through sketches
    for i,s in enumerate(all_sketches):
        if i%100==0:
            print 'Extracted {} of {} {} sketches...'.format(i, len(all_sketches), cat)
        if s in invalid:
            print s + ' marked invalid, moving on...'
        else:
            sketch_path = os.path.join(path_to_svg,c,s)
            # read in paths and loop through to get strokes and spline segments
            try:
                paths = read_svg_file(sketch_path)        
                for path in paths:
                    stroke_counter += 1
                    for point in path:
                        if type(point) == svgpathtools.path.CubicBezier:
                            start_x.append(np.real(point.start))
                            start_y.append(np.imag(point.start))   
                            c1_x.append(np.real(point.control1))
                            c1_y.append(np.imag(point.control1))
                            c2_x.append(np.real(point.control2))
                            c2_y.append(np.imag(point.control2))
                            end_x.append(np.real(point.end))
                            end_y.append(np.imag(point.end))
                            spline_num.append(spline_counter)
                            spline_counter += 1         
                            stroke_num.append(stroke_counter)
                            class_name.append(os.path.dirname(sketch_path).split('/')[-1])
                            photo_name.append(os.path.basename(sketch_path).split('-')[0] + '.jpg')
                            sketch_name.append(os.path.basename(sketch_path))   
            except:
                print 'Issue with ' + sketch_path + '... moving on.' 
                
    df = pd.DataFrame([start_x,start_y,c1_x,c1_y,c2_x,c2_y,end_x,end_y, \
                       stroke_num,spline_num,class_name,photo_name,sketch_name]) 
    df = df.transpose()
    df.columns = ['start_x','start_y','c1_x','c1_y','c2_x','c2_y','end_x','end_y', \
                  'stroke_num','spline_num','class_name','photo_name','sketch_name']

    if not os.path.exists(save_dir):
        os.makedirs(save_dir)
    save_path = os.path.join(save_dir, cat + '.csv')
    df.to_csv(save_path)


    

In [ ]:

    




    

In [111]:

    




    

In [128]:

    




    

In [ ]:

    




    

In [ ]: