def main(argv):
del argv
c = 0 # to count how many images it's gone through
s = 0 # to count how many images it's skipped
data = json.load(open(flags.ann_dir, 'r'))
processor = preprocess.Processor(data, flags.pickle, flags.is_training)
number_of_data = len(processor.selected_caption)
number_of_data_per_shard = number_of_data // flags.shards
save_dir = './train_tfrecords/' if flags.is_training else './eval_tfrecords/'
if not tf.gfile.IsDirectory(save_dir):
tf.gfile.MakeDirs(save_dir)
writers = [tf.io.TFRecordWriter(save_dir +'tfrecord_'+ str(i)) for i in range(flags.shards)]
print('Processing %d images' % number_of_data)
for idx, items in enumerate(zip(processor.selected_caption, processor.selected_image_id)):
caption, image_id = items
img_path = os.path.join(flags.img_dir, processor.img_idx2file[image_id])
try:
img = tf.io.gfile.GFile(img_path, 'rb').read()
except:
s += 1
print(img_path, "has been skipped")
continue
c += 1
converted_sentence = processor.converter(caption)
caption_data = processor.padding(converted_sentence)
features = {'image': _bytes_feature([img]), 'caption': _int_feature(caption_data)}
example = tf.train.Example(features=tf.train.Features(feature=features))
shard_idx = idx // number_of_data_per_shard
if shard_idx >= flags.shards:
shard_idx -= 1
writers[shard_idx].write(example.SerializeToString())
print('%d images in total' % c)
print('%d images have been skipped' % s)
def get_image_and_answer(annotation_path):
data = json.load(open(annotation_path, 'r'))
processor = preprocess.Processor(data, flags.pickle, False)
converted_captions = [
processor.converter(words) for words in processor.selected_caption
]
image_names = [
processor.img_idx2file[idx] for idx in processor.selected_image_id
]
img_name2answers = collections.defaultdict(list)
for img_name, answer in zip(image_names, converted_captions):
img_name2answers[img_name].append(answer)
img_names, answers = list(img_name2answers.keys()), list(
img_name2answers.values())
return img_names, answers
def startGUI():
cap = capture.Capture(0)
print('before run')
cap.start()
t = time.time()
while cap.img is None:
print('waiting for camera to connect',time.time()-t)
time.sleep(.25)
proc = preprocess.Processor()
proc.start(cap)
while proc.img is None:
print('waiting for processed image',time.time()-t)
time.sleep(.25)
ocr = OCR.OCR()
ocr.start(proc)
while ocr.txt is None:
print('waiting for OCR',time.time()-t)
time.sleep(.25)
root.cap = cap
root.proc = proc
root.mainloop()
from io import RawIOBase
import capture, preprocess, OCR
import time, cv2
cap = capture.Capture(1)
print('before run')
cap.start()
t = time.time()
while cap.img is None:
print('waiting for camera to connect',time.time()-t)
time.sleep(.25)
proc = preprocess.Processor()
proc.start(cap)
while proc.img is None:
print('waiting for processed image',time.time()-t)
time.sleep(.25)
ocr = OCR.OCR()
ocr.start(proc)
while ocr.txt is None:
print('waiting for OCR',time.time()-t)
time.sleep(.25)
t = time.time()
while True:
raw = cap.img