tf-image-segmentation.py

import os
import sys
import tensorflow as tf
import skimage.io as io
import skimage.morphology
import numpy as np

#sys.path.append('tf-image-segmentation')
sys.path.append('tf-image-segmentation-tf1.0')
sys.path.append('models/slim')
#sys.path.append('models-custom/slim')

fcn_16s_checkpoint_path = 'fcn_8s_checkpoint/model_fcn8s_final.ckpt'

# os.environ["CUDA_VISIBLE_DEVICES"] = '1'

slim = tf.contrib.slim

def save(image_np, pred_np):
    prediction_mask = (pred_np.squeeze() == 15)
    
    # Let's apply some morphological operations to
    # create the contour for our sticker
    
    cropped_object = image_np * np.dstack((prediction_mask,) * 3)
    
    square = skimage.morphology.square(5)
    
    temp = skimage.morphology.binary_erosion(prediction_mask, square)
    
    negative_mask = (temp != True)
    
    eroding_countour = negative_mask * prediction_mask
    
    eroding_countour_img = np.dstack((eroding_countour, ) * 3)
    
    cropped_object[eroding_countour_img] = 248
    
    png_transparancy_mask = np.uint8(prediction_mask * 255)
    
    image_shape = cropped_object.shape
    
    png_array = np.zeros(shape=[image_shape[0], image_shape[1], 4], dtype=np.uint8)
    
    png_array[:, :, :3] = cropped_object
    
    png_array[:, :, 3] = png_transparancy_mask
    
    io.imsave('sticker_me.png', png_array)

def main():
    from tf_image_segmentation.models.fcn_8s import FCN_8s
    from tf_image_segmentation.utils.inference import adapt_network_for_any_size_input
    from tf_image_segmentation.utils.pascal_voc import pascal_segmentation_lut

    number_of_classes = 21
    
    image_filename = 'me.jpg'
    
    image_filename_placeholder = tf.placeholder(tf.string)
    
    feed_dict_to_use = {image_filename_placeholder: image_filename}
    
    image_tensor = tf.read_file(image_filename_placeholder)
    
    image_tensor = tf.image.decode_jpeg(image_tensor, channels=3)
    
    image_batch_tensor = tf.expand_dims(image_tensor, axis=0)
    
    FCN_8s = adapt_network_for_any_size_input(FCN_8s, 32)
    
    pred, fcn_16s_variables_mapping = FCN_8s(image_batch_tensor=image_batch_tensor,
    					  number_of_classes=number_of_classes,
    					  is_training=False)
    
    initializer = tf.local_variables_initializer()
    
    saver = tf.train.Saver()
    
    with tf.Session() as sess:
        
        sess.run(initializer)
    
        saver.restore(sess, fcn_16s_checkpoint_path) 
        
        image_np, pred_np = sess.run([image_tensor, pred], feed_dict=feed_dict_to_use)
        
        io.imsave('image_np.png', image_np);
        io.imsave('pred_np.png', pred_np.squeeze())
    
        save(image_np, pred_np)

if __name__ == '__main__':
    main()