losses.py 文件源码

def pixel_loss(layer, FLAGS):
    generated_images, content_images = tf.split(0, 2, layer)

    #img_bytes = tf.read_file(FLAGS.mask_file)
    #maskimage = tf.image.decode_jpeg(img_bytes)
    #maskimage = tf.to_float(maskimage)
    #m_mean = tf.reduce_mean(maskimage, axis=(1,2))
    #index = tf.where(m_mean < 1.5)
    #top_index = index + tf.to_int64(1)
    #down_index = index - tf.to_int64(1)

    #select = tf.zeros_like(m_mean, dtype=tf.float32)
    #values = tf.squeeze(tf.ones_like(index, dtype=tf.float32))
    #topvalues = tf.squeeze(tf.ones_like(top_index, dtype=tf.float32))
    #downvalues = tf.squeeze(tf.ones_like(down_index, dtype=tf.float32))
    #delta = tf.SparseTensor(index, values, [FLAGS.image_size])
    #topdelta = tf.SparseTensor(index, topvalues, [FLAGS.image_size])
    #downdelta = tf.SparseTensor(index, downvalues, [FLAGS.image_size])
    #black_select = select + tf.sparse_tensor_to_dense(delta)
    #top_select = select + tf.sparse_tensor_to_dense(topdelta)
    #down_select = select + tf.sparse_tensor_to_dense(downdelta)

    #black_select = tf.mul(black_select, top_select)
    #black_select = tf.mul(black_select, down_select)
    #black_select = tf.expand_dims(black_select, -1)
    #black_select = tf.matmul(black_select, tf.ones([1, FLAGS.image_size]))
    #black_select = tf.expand_dims(black_select, -1)

    #generated_images = tf.mul(generated_images, black_select)
    #content_images = tf.mul(content_images, black_select)

    size = tf.size(generated_images)
    pixel_loss = tf.nn.l2_loss(generated_images - content_images) * 2 / tf.to_float(size)
    return pixel_loss