features/compute_features.py

'''
   File that computes features for a set of images

   ex. python compute_features.py --data_dir=/mnt/images/ --model=vgg19 --model_path=./vgg_19.ckpt

'''

import scipy.misc as misc
# import cPickle as pickle
import _pickle as cPickle
import tensorflow as tf
from tqdm import tqdm
import numpy as np
import argparse
import fnmatch
import sys
import os

sys.path.insert(0, 'nets/')

slim = tf.contrib.slim

'''

   Recursively obtains all images in the directory specified

'''
def getPaths(data_dir):  
   image_paths = []
   # add more extensions if need be
   ps = ['jpg', 'jpeg', 'JPG', 'JPEG', 'bmp', 'BMP', 'png', 'PNG']
   for p in ps:
      pattern   = '*.'+p
      for d, s, fList in os.walk(data_dir):
         for filename in fList:
            if fnmatch.fnmatch(filename, pattern):
               fname_ = os.path.join(d,filename)
               image_paths.append(fname_)
   return image_paths

if __name__ == '__main__':
   parser = argparse.ArgumentParser()
   parser.add_argument('--data_dir', required=True, type=str, help='Directory images are in. Searches recursively.')
   parser.add_argument('--model', required=True, type=str, help='Model to use')
   parser.add_argument('--checkpoint_file', required=True, type=str, help='Model file')
   a = parser.parse_args()

   data_dir        = a.data_dir
   model           = a.model
   checkpoint_file = a.checkpoint_file

   print( data_dir, model, checkpoint_file )

   # I only have these because I thought some take in size of (299,299), but maybe not
   if 'inception' in model: height, width, channels = 224, 224, 3
   if 'resnet' in model:    height, width, channels = 224, 224, 3
   if 'vgg' in model:       height, width, channels = 224, 224, 3

   if model == 'inception_resnet_v2': height, width, channels = 299, 299, 3

   x = tf.placeholder(tf.float32, shape=(1, height, width, channels))
   
   # load up model specific stuff
   if model == 'inception_v1':
      from inception_v1 import *
      arg_scope = inception_v1_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = inception_v1(x, is_training=False, num_classes=1001)
         features = end_points['AvgPool_0a_7x7']
   elif model == 'inception_v2':
      from inception_v2 import *
      arg_scope = inception_v2_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = inception_v2(x, is_training=False, num_classes=1001)
         features = end_points['AvgPool_1a']
   elif model == 'inception_v3':
      from inception_v3 import *
      arg_scope = inception_v3_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = inception_v3(x, is_training=False, num_classes=1001)
         features = end_points['AvgPool_1a']
   elif model == 'inception_resnet_v2':
      from inception_resnet_v2 import *
      arg_scope = inception_resnet_v2_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = inception_resnet_v2(x, is_training=False, num_classes=1001)
         features = end_points['PreLogitsFlatten']
   elif model == 'resnet_v1_50':
      from resnet_v1 import *
      arg_scope = resnet_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = resnet_v1_50(x, is_training=False, num_classes=1000)
         features = end_points['global_pool']
   elif model == 'resnet_v1_101':
      from resnet_v1 import *
      arg_scope = resnet_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = resnet_v1_101(x, is_training=False, num_classes=1000)
         features = end_points['global_pool']
   elif model == 'vgg_16':
      from vgg import *
      arg_scope = vgg_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = vgg_16(x, is_training=False)
         features = end_points['vgg_16/fc8']
   elif model == 'vgg_19':
      from vgg import *
      arg_scope = vgg_arg_scope()
      with slim.arg_scope(arg_scope):
         logits, end_points = vgg_19(x, is_training=False)
         features = end_points['vgg_19/fc8']

   print('init features...')
   sess  = tf.Session()
   saver = tf.train.Saver()
   saver.restore(sess, checkpoint_file)

   feat_dict = {}
   paths = getPaths(data_dir)
   print('Computing features...')
   for path in tqdm(paths):
      image = misc.imread(path)
      image = misc.imresize(image, (height, width))
      image = np.expand_dims(image, 0)
      feat = np.squeeze(sess.run(features, feed_dict={x:image}))
      feat_dict[path] = feat

   try: os.makedirs('features/')
   except: pass
   exp_pkl = open('features/'+model+'_features.pkl', 'wb')
   data = pickle.dumps(feat_dict)
   exp_pkl.write(data)
   exp_pkl.close()