def __init__(self, mode, max_epochs, batch_size, n_classes, train_records,
valid_records, im_sz, init_lr, keep_prob, logs_dir):
FCNNet.__init__(self, mode, max_epochs, batch_size, n_classes,
train_records, valid_records, im_sz, init_lr,
keep_prob, logs_dir)
self.cur_batch_size = tf.placeholder(dtype=tf.int32,
name='cur_batch_size')
#mask
self.seq_num = cfgs.seq_num
self.cur_channel = cfgs.cur_channel
self.channel = self.cur_channel + self.seq_num
self.inference_name = 'inference'
self.images = tf.placeholder(tf.float32,
shape=[
None, self.IMAGE_SIZE[0],
self.IMAGE_SIZE[1],
cfgs.seq_num + self.cur_channel
],
name='input_image')
self.create_view_path()
accu.create_ellipse_f()
self.e_acc = accu.Ellip_acc()
def __init__(self, mode, max_epochs, batch_size, n_classes, train_records, valid_records, im_sz, init_lr, keep_prob, logs_dir):
FCNNet.__init__(self, mode, max_epochs, batch_size, n_classes, train_records, valid_records, im_sz, init_lr, keep_prob, logs_dir)
#mask
self.seq_num = cfgs.seq_num
self.cur_channel = cfgs.cur_channel
self.channel = 3+self.seq_num
self.inference_name = 'soft_infer'
self.images = tf.placeholder(tf.float32, shape=[None, self.IMAGE_SIZE, self.IMAGE_SIZE, cfgs.seq_num+self.cur_channel], name='input_image')
accu.create_ellipse_f()
def __init__(self, mode, max_epochs, batch_size, n_classes, train_records,
valid_records, im_sz, init_lr, keep_prob, logs_dir):
FCNNet.__init__(self, mode, max_epochs, batch_size, n_classes,
train_records, valid_records, im_sz, init_lr,
keep_prob, logs_dir)
#mask
self.seq_num = cfgs.seq_num
self.cur_channel = cfgs.cur_channel
self.channel = self.cur_channel + self.seq_num
self.inference_name = 'inference'
self.images = tf.placeholder(tf.float32,
shape=[
None, self.IMAGE_SIZE[0],
self.IMAGE_SIZE[1],
cfgs.seq_num + self.cur_channel
],
name='input_image')
self.grad_ims = tf.placeholder(
tf.float32,
shape=[None, self.IMAGE_SIZE[0], self.IMAGE_SIZE[1], 1],
name='grad_image')
self.create_view_path()
self.coord_map_x, self.coord_map_y = self.generate_coord_map(
self.batch_size)
self.coord_x_tensor = tf.placeholder(
tf.float32,
shape=[None, self.IMAGE_SIZE[0], self.IMAGE_SIZE[1]],
name='coord_x_map_tensor')
self.coord_y_tensor = tf.placeholder(
tf.float32,
shape=[None, self.IMAGE_SIZE[0], self.IMAGE_SIZE[1]],
name='coord_y_map_tensor')
self.ellip_low = tf.placeholder(tf.float32,
shape=[None],
name='ellipse_info_lower_axis')
self.ellip_high = tf.placeholder(tf.float32,
shape=[None],
name='ellipse_info_higher_axis')
self.ellip_axis = tf.placeholder(tf.float32,
shape=[None],
name='ellipse_info_mean_axis')
accu.create_ellipse_f()
self.e_acc = accu.Ellip_acc()
def __init__(self, mode, max_epochs, batch_size, n_classes, train_records,
valid_records, im_sz, init_lr, keep_prob, logs_dir):
FCNNet.__init__(self, mode, max_epochs, batch_size, n_classes,
train_records, valid_records, im_sz, init_lr,
keep_prob, logs_dir)
#seq_mask(short for sm)
self.seq_num = cfgs.seq_num
self.cur_channel = cfgs.cur_channel
self.sm_channel = self.cur_channel + self.seq_num
self.sm_infer_name = 'inference'
self.sm_images = tf.placeholder(tf.float32,
shape=[
None, self.IMAGE_SIZE,
self.IMAGE_SIZE,
cfgs.seq_num + self.cur_channel
],
name='seq_mask_input_image')
self.sm_annos = tf.placeholder(tf.int32,
shape=[
None, self.IMAGE_SIZE,
self.IMAGE_SIZE,
cfgs.seq_mask_anno_channel
],
name='seq_mask_annos')
#soft
self.soft_infer_name = 'soft_infer'
self.soft_channel = 3
self.soft_anno_channel = cfgs.soft_anno_channel
self.soft_images = tf.placeholder(
tf.float32,
shape=[None, self.IMAGE_SIZE, self.IMAGE_SIZE, self.soft_channel],
name='soft_input_images')
self.soft_annos = tf.placeholder(tf.float32,
shape=[
None, self.IMAGE_SIZE,
self.IMAGE_SIZE,
self.soft_anno_channel
],
name='soft_annos')
accu.create_ellipse_f()
self.create_view_f()
self.e_acc = accu.Ellip_acc()