Bug: Using images out of ImageSets\Main\trainval.txt for training

[jamiechoi1995]

In convert_tfrecords.py, it converts all images in VOCdevkit\VOC2012\JPEGImages into tfrecords for training,

SSD.TensorFlow/dataset/convert_tfrecords.py

Lines 366 to 370 in b47ff61

	for split in all_splits:
	jpeg_file_path = os.path.join(directory, split, 'JPEGImages')
	images = tf.gfile.ListDirectory(jpeg_file_path)
	jpegs = [im_name for im_name in images if im_name.strip()[-3:]=='jpg']
	all_records.extend(list(zip([split] * len(jpegs), jpegs)))

However, not all the images in VOC2012 are for detection, there are only 11540 images specified in ImageSets\Main\trainval.txt for detection. But in JPEGImages there are 17125 images, these extra images are for action classification, so they only annotate human, ignoring other classes.
Eg,

Using these images will be harmful for training.

[jamiechoi1995]

after removing the images and using 11540 images to train the model, I got 0.7785 mAP with the default settings using model-ckpt-116604