Merge pull request hunglc007#52 from wooruang/master

Add converter for voc

scripts/voc/README.md

@@ -0,0 +1,27 @@
+# VOC Dataset
+
+### Download
+
+```bash
+$ bash get_voc2012.sh
+```
+
+### Make names for VOC.
+
+```bash
+$ python voc_make_names.py [--anno_dir {Annotation directory}] [--output {OUTPUT_NAME}]
+
+# example
+$ python voc_make_name.py
+
+$ python voc_make_name.py --anno_dir ../../data/voc/anno --output ../../data/classes/voc.names
+```
+
+### Convert VOC Dataset.
+
+```bash
+$ python voc_convert.py [--image_dir {Image directory}] [--anno_dir {Annotation directory}] [--train_list_txt {Path of Train list file}] [--val_list_txt {Path of Validation list file}] [--classes {Path of Classes file}] [--train_output {Path of Output file For Train}] [--val_output {Path of Output file For Val}]
+
+#example
+$ python voc_convert.py
+```

scripts/voc/get_voc2012.sh

@@ -0,0 +1,19 @@
+#!/usr/bin/env bash
+
+
+PARENT_DIR="$( cd "$( dirname $( dirname $( dirname "${BASH_SOURCE[0]}" ) ) )" >/dev/null 2>&1 && pwd )"
+DATA_DIR="$PARENT_DIR/data"
+
+DATASET_NAME="VOCtrainval_11-May-2012"
+
+wget -c -P $DATA_DIR http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
+
+if [[ -d "$DATA_DIR/$DATASET_NAME" ]]; then
+    echo "Already '$DATA_DIR/$DATASET_NAME' path exists."
+    exit 1
+fi
+
+mkdir $DATA_DIR/$DATASET_NAME
+
+tar xf $DATA_DIR/VOCtrainval_11-May-2012.tar -C $DATA_DIR/$DATASET_NAME
+

scripts/voc/voc_convert.py

@@ -0,0 +1,78 @@
+import sys
+import os
+
+from absl import app, flags
+from absl.flags import FLAGS
+from lxml import etree
+
+
+flags.DEFINE_string('image_dir', '../../data/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/JPEGImages', 'path to image dir')
+flags.DEFINE_string('anno_dir', '../../data/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/Annotations', 'path to anno dir')
+flags.DEFINE_string('train_list_txt', '../../data/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/train.txt', 'path to a set of train')
+flags.DEFINE_string('val_list_txt', '../../data/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/ImageSets/Main/val.txt', 'path to a set of val')
+flags.DEFINE_string('classes', '../../data/classes/voc2012.names', 'path to a list of class names')
+flags.DEFINE_string('train_output', '../../data/dataset/voc2012_train.txt', 'path to a file for train')
+flags.DEFINE_string('val_output', '../../data/dataset/voc2012_val.txt', 'path to a file for val')
+
+flags.DEFINE_boolean('no_val', False, 'if uses this flag, it does not convert a list of val')
+
+
+def convert_annotation(list_txt, output_path, image_dir, anno_dir, class_names):
+    IMAGE_EXT = '.jpg'
+    ANNO_EXT = '.xml'
+
+    with open(list_txt, 'r') as f, open(output_path, 'w') as wf:
+        while True:
+            line = f.readline().strip()
+            if line is None or not line:
+                break
+            im_p = os.path.join(image_dir, line + IMAGE_EXT)
+            an_p = os.path.join(anno_dir, line + ANNO_EXT)
+
+            # Get annotation.
+            root = etree.parse(an_p).getroot()
+            bboxes = root.xpath('//object/bndbox')
+            names = root.xpath('//object/name')
+
+            box_annotations = []
+            for b, n in zip(bboxes, names):
+                name = n.text
+                class_idx = class_names.index(name)
+
+                xmin = b.find('xmin').text
+                ymin = b.find('ymin').text
+                xmax = b.find('xmax').text
+                ymax = b.find('ymax').text
+                box_annotations.append(','.join([str(xmin), str(ymin), str(xmax), str(ymax), str(class_idx)]))
+
+            annotation = os.path.abspath(im_p) + ' ' + ' '.join(box_annotations) + '\n'
+
+            wf.write(annotation)
+
+
+def convert_voc(image_dir, anno_dir, train_list_txt, val_list_txt, classes, train_output, val_output, no_val):
+    IMAGE_EXT = '.jpg'
+    ANNO_EXT = '.xml'
+
+    class_names = [c.strip() for c in open(FLAGS.classes).readlines()]
+
+    # Training set.
+    convert_annotation(train_list_txt, train_output, image_dir, anno_dir, class_names)
+
+    if no_val:
+        return
+
+    # Validation set.
+    convert_annotation(val_list_txt, val_output, image_dir, anno_dir, class_names)
+
+
+def main(_argv):
+    convert_voc(FLAGS.image_dir, FLAGS.anno_dir, FLAGS.train_list_txt, FLAGS.val_list_txt, FLAGS.classes, FLAGS.train_output, FLAGS.val_output, FLAGS.no_val)
+    print("Complete convert voc data!")
+
+
+if __name__ == "__main__":
+    try:
+        app.run(main)
+    except SystemExit:
+        pass    

scripts/voc/voc_make_names.py

@@ -0,0 +1,46 @@
+import sys
+import os
+
+from absl import app, flags
+from absl.flags import FLAGS
+from lxml import etree
+
+
+flags.DEFINE_string('anno_dir', '../../data/VOCtrainval_11-May-2012/VOCdevkit/VOC2012/Annotations', 'path to anno dir')
+flags.DEFINE_string('output', '../../data/classes/voc2012.names', 'path to anno dir')
+
+
+def make_names(anno_dir, output):
+    labels_dict = {}
+
+    anno_list = os.listdir(anno_dir)
+
+    for anno_file in anno_list:
+        p = os.path.join(anno_dir, anno_file)
+
+        # Get annotation.
+        root = etree.parse(p).getroot()
+        names = root.xpath('//object/name')
+
+        for n in names:
+            labels_dict[n.text] = 0
+
+    labels = list(labels_dict.keys())
+    labels.sort()
+
+    with open(output, 'w') as f:
+        for l in labels:
+            f.writelines(l + '\n')
+
+    print(f"Done making a names's file ({os.path.abspath(output)})")
+
+
+def main(_argv):
+    make_names(FLAGS.anno_dir, FLAGS.output)
+
+
+if __name__ == "__main__":
+    try:
+        app.run(main)
+    except SystemExit:
+        pass