sess.run(init)
# restore previous model if there is one
ckpt = tf.train.get_checkpoint_state(CHECKPOINT_DIR)
if ckpt and ckpt.model_checkpoint_path:
print "Restoring previous model..."
try:
saver.restore(sess, ckpt.model_checkpoint_path)
print "Model restored"
except:
print "Could not restore model"
raise
exit()
print 'Loading data...'
train_images, train_annots, train_ids, test_images, test_annots, test_ids = data_ops.load_efigi(
DATA_DIR, CLASSES, 64)
test_len = len(test_ids)
print 'generating data...'
idx = np.random.choice(np.arange(test_len), BATCH_SIZE, replace=False)
batch_z = np.random.normal(-1.0, 1.0, size=[BATCH_SIZE,
100]).astype(np.float32)
batch_z[1:] = batch_z[0]
batch_y = test_annots[idx]
batch_y[:NUM +
1] = batch_y[1] # gotta make sure they have the same attributes
# making up my own
#batch_y[0][0] = 1. # arm strength
#batch_y[0][1] = 1. # arm curvature
#batch_y[0][2] = 0. # visible dust
exp_info = dict()
exp_info['CHECKPOINT_DIR'] = CHECKPOINT_DIR
exp_info['BATCH_SIZE'] = BATCH_SIZE
exp_info['DATA_DIR'] = DATA_DIR
exp_info['CLASSES'] = classes
exp_info['EPOCHS'] = EPOCHS
exp_info['LOSS'] = LOSS
exp_pkl = open(CHECKPOINT_DIR + 'info.pkl', 'wb')
data = pickle.dumps(exp_info)
exp_pkl.write(data)
exp_pkl.close()
print 'Loading data...'
train_images, train_annots, train_ids, test_images, test_annots, test_ids = data_ops.load_efigi(
DATA_DIR, classes, 64)
# find y dimension
idx_ = np.array(
[1, 4, 7, 10, 13, 16, 19, 22, 25, 28, 31, 34, 37, 40, 43, 46, 49])
idx_ = np.multiply(classes[:-1], idx_)
idx = [x for x in idx_ if x != 0]
y_dim = len(idx)
# account for redshift attribute
if classes[-1] == 1: y_dim += 1
# placeholders for data going into the network
global_step = tf.Variable(0, name='global_step', trainable=False)
real_images = tf.placeholder(tf.float32,
shape=(BATCH_SIZE, 64, 64, 3),
name='real_images')