Merge pull request kaldi-asr#15 from aarora8/madcat_ar_2

Madcat ar 2

egs/madcat_ar/v1/local/create_line_image_from_page_image.py

@@ -22,6 +22,7 @@
 from math import atan2, cos, sin, pi, degrees, sqrt
 from collections import namedtuple
 
+from scipy.spatial import ConvexHull
 from PIL import Image
 from scipy.misc import toimage
 import logging
@@ -262,7 +263,10 @@ def minimum_bounding_box(points):
     """
 
     if len(points) <= 2: raise ValueError('More than two points required.')
-    hull_ordered = compute_hull(points)
+
+    hull_ordered = [points[index] for index in ConvexHull(points).vertices]
+    hull_ordered.append(hull_ordered[0])
+    #hull_ordered = compute_hull(points)
     hull_ordered = tuple(hull_ordered)
 
     min_rectangle = bounding_area(0, hull_ordered)
@@ -406,24 +410,9 @@ def set_line_image_data(image, line_id, image_file_name, image_fh):
     image_fh.write(image_path + '\n')
 
 
-def paste_line_image_on_blank_image(im, line_image, height_offset, width_offset):
-    """ Given line image and its location on page image, return a page image
-        with only line image in ti.
-        It is done to remove noise that can occur due to MAR. 
-    :return: 
-    --------
-    image: page image containing only line image
-    """
-    image_width= im.size[0]
-    image_height = im.size[1]
-    stitched_image = Image.new('L', (image_width, image_height), "white")
-    stitched_image.paste(im=line_image, box=(width_offset, height_offset))
-    return stitched_image
-
-
 def get_line_images_from_page_image(image_file_name, madcat_file_path, image_fh):
     """ Given a page image, extracts the line images from it.
-    Input
+    Inout
     -----
     image_file_name (string): complete path and name of the page image.
     madcat_file_path (string): complete path and name of the madcat xml file
@@ -433,22 +422,8 @@ def get_line_images_from_page_image(image_file_name, madcat_file_path, image_fh)
     im = pad_image(im_wo_pad)
     doc = minidom.parse(madcat_file_path)
     zone = doc.getElementsByTagName('zone')
-    offset = int(args.padding // 2)
     for node in zone:
         id = node.getAttribute('id')
-        line_point = node.getElementsByTagName('point')
-        col, row = [], []
-        max_col, max_row, min_col, min_row = 0, 0, 0, 0
-        for line_node in line_point:
-            col.append(int(line_node.getAttribute('x')) + offset)
-            row.append(int(line_node.getAttribute('y')) + offset)
-            max_col, max_row = max(col), max(row)
-            min_col, min_row = min(col), min(row)
-
-        box = (min_col, min_row, max_col, max_row)
-        region = im.crop(box)
-        image = paste_line_image_on_blank_image(im, region, min_row, min_col)
-
         token_image = node.getElementsByTagName('token-image')
         minimum_bounding_box_input = []
         for token_node in token_image:
@@ -469,7 +444,7 @@ def get_line_images_from_page_image(image_file_name, madcat_file_path, image_fh)
         max_x = int(max(x1, x2, x3, x4))
         max_y = int(max(y1, y2, y3, y4))
         box = (min_x, min_y, max_x, max_y)
-        region_initial = image.crop(box)
+        region_initial = im.crop(box)
         rot_points = []
         p1_new = (x1 - min_x, y1 - min_y)
         p2_new = (x2 - min_x, y2 - min_y)
@@ -603,3 +578,4 @@ def main():
 
 if __name__ == '__main__':
       main()
+

egs/madcat_ar/v1/local/extract_features.sh

@@ -31,7 +31,7 @@ utils/split_scp.pl $scp $split_scps || exit 1;
 
 $cmd JOB=1:$nj $logdir/extract_features.JOB.log \
   local/make_features.py $logdir/images.JOB.scp \
-    --allowed_len_file_path data/train \
+    --allowed_len_file_path $data/allowed_lengths.txt \
     --feat-dim $feat_dim \| \
     copy-feats --compress=true --compression-method=7 \
     ark:- ark,scp:$featdir/images.JOB.ark,$featdir/images.JOB.scp

egs/madcat_ar/v1/local/make_features.py

@@ -105,7 +105,7 @@ def horizontal_pad(im, allowed_lengths = None):
     out_fh = open(args.out_ark,'wb')
 
 allowed_lengths = None
-allowed_len_handle = '/'.join(filter(None, (args.allowed_len_file_path, 'allowed_lengths.txt')))
+allowed_len_handle = args.allowed_len_file_path
 if os.path.isfile(allowed_len_handle):
     print("Found 'allowed_lengths.txt' file...", file=sys.stderr)
     allowed_lengths = []

egs/madcat_ar/v1/local/wer_output_filter

@@ -12,7 +12,7 @@ binmode STDERR, ":utf8";
 # Arabic-specific normalization
 while (<>) {
   @F = split " ";
-  print $F[0] . " "; 
+  print $F[0] . " ";
   foreach $s (@F[1..$#F]) {
     $s =~ s/\x{0623}/\x{0627}/g;
     $s =~ s/\x{0625}/\x{0627}/g;