def get_mynet_input(key):
class mynetRecord(object):
pass
result = mynetRecord()
label_bytes = 4
result.height = 64
result.width = 64
result.depth = 4
print(key)
if (key[0] == 'shape'):
target = target_gen.generate_image(requested_shape=key[1],
requested_label=key[0])
elif (key[0] == 'letter'):
target = target_gen.generate_image(requested_letter=key[1],
requested_label=key[0])
result.label = target.label
print("label: " + str(result.label))
record_bytes = tf.decode_raw(target.image, tf.uint8)
depth_major = tf.reshape(record_bytes,
[result.depth, result.height, result.width])
result.uint8image = tf.transpose(depth_major, [1, 2, 0])
return result
import target_gen
import sys
import argparse
parser = argparse.ArgumentParser(
description='request specific shape,letter,colors')
parser.add_argument('-s', help='generate a specific shape')
parser.add_argument('-l', help='generate a specific letter')
parser.add_argument('-sc', help='generate a specific shape color')
parser.add_argument('-lc', help='generate a specific letter color')
args = parser.parse_args()
s = l = sc = lc = None
if (args.s):
s = args.s
if (args.l):
l = args.l
if (args.sc):
sc = args.sc
if (args.lc):
lc = args.lc
print(
target_gen.generate_image(requested_shape=s,
requested_letter=l,
requested_shape_color=sc,
requested_letter_color=lc))
labels = [None] * num_training_images
images_test = [None] * num_testing_images
labels_test = [None] * num_testing_images
letter_list = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z']
shape_list = ['Circle', 'Semicircle', 'Quartercircle', 'Triangle', 'Square', 'Rectangle', 'Trapezoid', 'Pentagon', 'Hexagon', 'Heptagon', 'Octagon', 'Star', 'Cross']
counter = 0
for i in range(0, 26):
for a in range(0, 13):
tmp_img, tmp_label = target_gen.generate_image(requested_letter=letter_list[i],
requested_shape=shape_list[a],
requested_letter_color="White",
requested_shape_color="Black",
return_type = "set")
for q in range(0, num_variations):
tmp_img_2 = tmp_img
tmp_img_2 = tmp_img_2.filter(ImageFilter.FIND_EDGES)
tmp_img_2 = tmp_img_2.filter(ImageFilter.SMOOTH)
tmp_img_2 = tmp_img_2.filter(ImageFilter.SMOOTH_MORE)
tmp_img_2 = tmp_img_2.filter(ImageFilter.FIND_EDGES)
tmp_img_2 = tmp_img_2.convert('L')
tmp_img_2 = tmp_img_2.filter(ImageFilter.EDGE_ENHANCE_MORE)
images[counter] = np.reshape(tmp_img_2.getdata(), (64, 64, -1))
#labels[counter] = np.reshape(tmp_label, (-1))
#labels[counter] = tflearn.data_utils.to_categorical(tmp_label,338)
labels[counter] = tmp_label
import tflearn
import numpy as np
from tflearn.data_utils import shuffle
from tflearn.data_utils import *
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d
from tflearn.layers.estimator import regression
from tflearn.data_preprocessing import ImagePreprocessing
from tflearn.data_augmentation import ImageAugmentation
import pickle
import target_gen
from PIL import Image
from PIL import ImageFilter
import sys
np.set_printoptions(threshold=np.inf)
# load dataset of auvsi targets
# or generate them on demand here??
num_testing_images = 100
if sys.argv[1]:
num_testing_images = int(sys.argv[1])
for i in range(0, num_testing_images):
tmp_img, tmp_label = target_gen.generate_image(return_type="shape")
sys.stdout.write("Generating image %d/%d \r" % (i, num_testing_images))
sys.stdout.flush()
sys.stdout.write("Generating image %d/%d \n" % (i + 1, num_testing_images))
sys.stdout.write("Finished generating images\n")