1st

PSNR.py

@@ -0,0 +1,14 @@
+
+import numpy as np
+import math
+
+def psnr(target, ref):
+	#assume RGB image
+	target_data = np.array(target)
+
+	ref_data = np.array(ref)
+
+	diff = ref_data - target_data
+	diff = diff.flatten('C')
+	rmse = math.sqrt( np.mean(diff ** 2.) )
+	return 20*math.log10(1.0/rmse)

enhancenet.py

@@ -0,0 +1,44 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+from nnlib import *
+import PSNR
+
+PER_CHANNEL_MEANS = np.array([0.47614917, 0.45001204, 0.40904046])
+fns = sorted([fn for fn in os.listdir('input/')])
+if not os.path.exists('output'):
+    os.makedirs('output')
+for fn in fns:
+    fne = ''.join(fn.split('.')[:-1])
+    if os.path.isfile('output/%s-EnhanceNet.png' % fne):
+        print('skipping %s' % fn)
+        continue
+    imgs = loadimg('input/'+fn)
+    if imgs is None:
+        continue
+    imgs = np.expand_dims(imgs, axis=0)
+    imgsize = np.shape(imgs)[1:]
+    print('processing %s' % fn)
+    xs = tf.placeholder(tf.float32, [1, imgsize[0], imgsize[1], imgsize[2]])
+    rblock = [resi, [[conv], [relu], [conv]]]
+    ys_est = NN('generator',
+                [xs,
+                 [conv], [relu],
+                 rblock, rblock, rblock, rblock, rblock,
+                 rblock, rblock, rblock, rblock, rblock,
+                 [upsample], [conv], [relu],
+                 [upsample], [conv], [relu],
+                 [conv], [relu],
+                 [conv, 3]])
+    ys_res = tf.image.resize_images(xs, [4*imgsize[0], 4*imgsize[1]],
+                                    method=tf.image.ResizeMethod.BICUBIC)
+    ys_est += ys_res + PER_CHANNEL_MEANS
+    sess = tf.InteractiveSession()
+    tf.train.Saver().restore(sess, os.getcwd()+'/weights')
+    output = sess.run([ys_est, ys_res+PER_CHANNEL_MEANS],
+                      feed_dict={xs: imgs-PER_CHANNEL_MEANS})
+
+    #print(PSNR.psnr(output[0][0][::4,::4,:], imgs))
+    saveimg(output[0][0], 'output/%s-EnhanceNet.png' % fne)
+    saveimg(output[1][0], 'output/%s-Bicubic.png' % fne)
+    sess.close()
+    tf.reset_default_graph()

eval.ipynb

@@ -0,0 +1 @@
+{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"eval.ipynb","version":"0.3.2","views":{},"default_view":{},"provenance":[],"collapsed_sections":[]},"kernelspec":{"name":"python3","display_name":"Python 3"},"accelerator":"GPU"},"cells":[{"metadata":{"id":"oMKxe3FlTO9Z","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0},"base_uri":"https://localhost:8080/","height":105},"outputId":"fcb4e363-52af-4c25-81cf-27e0a48f117b","executionInfo":{"status":"ok","timestamp":1525861235345,"user_tz":-330,"elapsed":20596,"user":{"displayName":"Ravi Tej Akella","photoUrl":"//lh4.googleusercontent.com/-CxyhmYRONEI/AAAAAAAAAAI/AAAAAAAAR2A/PCmQzRxsHmo/s50-c-k-no/photo.jpg","userId":"106703338527093367257"}}},"cell_type":"code","source":["!apt-get install -y -qq software-properties-common python-software-properties module-init-tools\n","!add-apt-repository -y ppa:alessandro-strada/ppa 2>&1 > /dev/null\n","!apt-get update -qq 2>&1 > /dev/null\n","!apt-get -y install -qq google-drive-ocamlfuse fuse\n","from google.colab import auth\n","auth.authenticate_user()\n","from oauth2client.client import GoogleCredentials\n","creds = GoogleCredentials.get_application_default()\n","import getpass\n","!google-drive-ocamlfuse -headless -id={creds.client_id} -secret={creds.client_secret} < /dev/null 2>&1 | grep URL\n","vcode = getpass.getpass()\n","!echo {vcode} | google-drive-ocamlfuse -headless -id={creds.client_id} -secret={creds.client_secret}"],"execution_count":1,"outputs":[{"output_type":"stream","text":["Please, open the following URL in a web browser: https://accounts.google.com/o/oauth2/auth?client_id=32555940559.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive&response_type=code&access_type=offline&approval_prompt=force\r\n","··········\n","Please, open the following URL in a web browser: https://accounts.google.com/o/oauth2/auth?client_id=32555940559.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive&response_type=code&access_type=offline&approval_prompt=force\n","Please enter the verification code: Access token retrieved correctly.\n"],"name":"stdout"}]},{"metadata":{"id":"HDA8JfxgTUGk","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0}}},"cell_type":"code","source":["!mkdir -p drive\n","!google-drive-ocamlfuse drive"],"execution_count":0,"outputs":[]},{"metadata":{"id":"0EjTxU_36dbq","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0},"base_uri":"https://localhost:8080/","height":394},"outputId":"af8ad33b-ea04-41d9-b2f8-b99a7eb679fa","executionInfo":{"status":"ok","timestamp":1525861371344,"user_tz":-330,"elapsed":122732,"user":{"displayName":"Ravi Tej Akella","photoUrl":"//lh4.googleusercontent.com/-CxyhmYRONEI/AAAAAAAAAAI/AAAAAAAAR2A/PCmQzRxsHmo/s50-c-k-no/photo.jpg","userId":"106703338527093367257"}}},"cell_type":"code","source":["import os\n","#!ls\n","os.chdir(\"/content/drive/App/Enhancenet\")\n","!python enhancenet.py\n"],"execution_count":3,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\r\n","  from ._conv import register_converters as _register_converters\n","processing baboon.bmp\n","processing barbara.bmp\n","processing bridge.bmp\n","processing coastguard.bmp\n","processing comic.bmp\n","processing face.bmp\n","processing flowers.bmp\n","processing foreman.bmp\n","processing lenna.bmp\n","processing man.bmp\n","processing monarch.bmp\n","processing pepper.bmp\n","processing sun_aaalbzqrimafwbiv.jpg\n","processing sun_aaaulhwrhqgejnyt.jpg\n","processing sun_aacphuqehdodwawg.jpg\n","processing sun_aacyknxirsfolpon.jpg\n","processing sun_aakfpvtynorwesef.jpg\n","processing sun_aamvxnvouicstkjb.jpg\n","processing zebra.bmp\n"],"name":"stdout"}]},{"metadata":{"id":"cPmg_utaXcE8","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0},"base_uri":"https://localhost:8080/","height":88},"outputId":"75b198d6-926b-4aa6-945d-2eb244c5b32b","executionInfo":{"status":"ok","timestamp":1525833657681,"user_tz":-330,"elapsed":81298,"user":{"displayName":"Ravi Tej Akella","photoUrl":"//lh4.googleusercontent.com/-CxyhmYRONEI/AAAAAAAAAAI/AAAAAAAAR2A/PCmQzRxsHmo/s50-c-k-no/photo.jpg","userId":"106703338527093367257"}}},"cell_type":"code","source":["'''%matplotlib inline\n","import matplotlib.pyplot as plt\n","import os\n","os.chdir(\"/content/drive/App/EnhanceNet\")\n","!python mod_enhancenet.py\n","# upscales HR image by 4x zoom'''"],"execution_count":4,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\r\n","  from ._conv import register_converters as _register_converters\n","(3, 256, 256, 3)\n"],"name":"stdout"}]},{"metadata":{"id":"ivXqnhdVGlb_","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0},"base_uri":"https://localhost:8080/","height":173},"outputId":"9e45b039-0319-4f36-b780-3ee32edb32ee","executionInfo":{"status":"ok","timestamp":1525820031786,"user_tz":-330,"elapsed":25797,"user":{"displayName":"Ravi Tej Akella","photoUrl":"//lh4.googleusercontent.com/-CxyhmYRONEI/AAAAAAAAAAI/AAAAAAAAR2A/PCmQzRxsHmo/s50-c-k-no/photo.jpg","userId":"106703338527093367257"}}},"cell_type":"code","source":["'''%matplotlib inline\n","import matplotlib.pyplot as plt\n","import os\n","os.chdir(\"/content/drive/App/EnhanceNet\")\n","!python modified_enhancenet.py\n","# upscales LR image by 4x zoom\n","######@@@@@@@@@@@@@ OOM problem in enhance graph for 256x256 input'''"],"execution_count":4,"outputs":[{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\r\n","  from ._conv import register_converters as _register_converters\n","2018-05-08 22:54:15.749260: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 528.50MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n","2018-05-08 22:54:15.768065: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.05GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n","2018-05-08 22:54:15.769560: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 288.56MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n","2018-05-08 22:54:15.783967: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 144.00MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n","2018-05-08 22:54:15.789339: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.12GiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n","2018-05-08 22:54:15.817047: W tensorflow/core/common_runtime/bfc_allocator.cc:219] Allocator (GPU_0_bfc) ran out of memory trying to allocate 603.03MiB. The caller indicates that this is not a failure, but may mean that there could be performance gains if more memory were available.\n"],"name":"stdout"}]},{"metadata":{"id":"DUD_-NI8xdDk","colab_type":"code","colab":{"autoexec":{"startup":false,"wait_interval":0}}},"cell_type":"code","source":[""],"execution_count":0,"outputs":[]}]}

mod_enhancenet.py

@@ -0,0 +1,119 @@
+#!/usr/bin/env python
+# -*- coding: UTF-8 -*-
+from nnlib import *
+#%matplotlib inline
+
+import numpy as np
+import tensorflow as tf
+import matplotlib.pyplot as plt
+import os
+os.chdir("/content/drive/App/EnhanceNet")
+
+#import cv2
+from PIL import Image
+image_files = os.listdir("/content/drive/App/CVAE/natural_images")
+trainX = []
+# print(len(image_files))
+for i, file in enumerate(image_files):
+  if i%100 == 0 and file.endswith(".jpg"):
+    img = Image.open("/content/drive/App/CVAE/natural_images/"+file).convert('RGB')
+    w, h = img.size
+#   ar = cv2.cvtColor(cv2.imread("/content/drive/App/CVAE/natural_images/"+file),cv2.COLOR_BGR2RGB)
+    trainX.append(np.array(img)/255)
+trainX = np.asarray(trainX)
+#trainX = trainX/255
+#plt.axis("off")
+#plt.imshow(trainX[0])
+
+from sklearn.model_selection import train_test_split
+x_train, x_test = train_test_split(trainX, test_size = 0.1, random_state=2)
+
+###############################################################################################
+import tensorflow.contrib.layers as lays
+def autoencoder(inputs):
+  with tf.variable_scope("CAE", reuse = tf.AUTO_REUSE) :
+    net = lays.conv2d(inputs, 3, [3, 3], stride=2, padding='SAME')
+    net = lays.conv2d(net, 3, [3, 3], stride=2, padding='SAME')
+    net0 = lays.conv2d(net, 3, [3, 3], stride=2, padding='SAME')
+    #net = lays.conv2d(net, 3, [3, 3], stride=2, padding='SAME')
+
+    upsampled256 = tf.image.resize_bicubic(net, (256,256))
+    upsampled128 = tf.image.resize_bicubic(net, (128,128))
+    upsampled64 = tf.image.resize_bicubic(net, (64,64))
+
+    net1 = lays.conv2d_transpose(net0, 3, [3, 3], stride=2, padding='SAME')
+    net1 = tf.add(net1,upsampled64)
+    net2 = lays.conv2d_transpose(net1, 3, [3, 3], stride=2, padding='SAME')
+    net2 = tf.add(net2,upsampled128)
+    net3 = lays.conv2d_transpose(net2, 3, [3, 3], stride=2, padding='SAME', activation_fn=tf.nn.sigmoid)
+    net3 = tf.add(net3,upsampled256)
+  return [net3,net1]
+
+ae_inputs = tf.placeholder(tf.float32, (None, 256, 256, 3))
+ae_outputs,net1 = autoencoder(ae_inputs)       #tf.image.resize_bicubic(temp, (256,256))
+#loss = tf.reduce_mean(tf.square(ae_outputs - ae_inputs))
+#train_op = tf.train.AdamOptimizer(learning_rate=lr).minimize(loss)
+init = tf.global_variables_initializer()
+#################################
+'''all_vars = tf.global_variables()
+model_one_vars = [k for k in all_vars if k.name.startswith("CAE")]
+print(len(all_vars))
+print(len(model_one_vars))'''
+#################################
+saver = tf.train.Saver()
+
+with tf.Session() as sess:
+  sess.run(init)
+  saver.restore(sess, "/content/drive/App/EnhanceNet/models/model.ckpt")
+  #batch_img = x_test
+  recon_LR_img, recon_HR_img = sess.run([net1, ae_outputs], feed_dict={ae_inputs: x_test})
+###############################################################################################
+timepas = recon_HR_img
+print(timepas.shape)
+PER_CHANNEL_MEANS = np.array([0.47614917, 0.45001204, 0.40904046])
+for i, image in enumerate(recon_HR_img[:5]) :
+    imgs = np.expand_dims(image, axis=0)
+    imgsize = np.shape(imgs)[1:]
+    #print('processing %s' % fn)
+    xs = tf.placeholder(tf.float32, [1, imgsize[0], imgsize[1], imgsize[2]])
+    rblock = [resi, [[conv], [relu], [conv]]]
+    ys_est = NN('generator',
+                [xs,
+                 [conv], [relu],
+                 rblock, rblock, rblock, rblock, rblock,
+                 rblock, rblock, rblock, rblock, rblock,
+                 [upsample], [conv], [relu],
+                 [upsample], [conv], [relu],
+                 [conv], [relu],
+                 [conv, 3]])
+    ys_res = tf.image.resize_images(xs, [4*imgsize[0], 4*imgsize[1]],
+                                    method=tf.image.ResizeMethod.BICUBIC)
+    ys_est += ys_res + PER_CHANNEL_MEANS
+    sess = tf.InteractiveSession()
+    ###################################################################
+    all_vars = tf.global_variables()
+    model_one_vars = [k for k in all_vars if k.name.startswith("CAE")]
+    set_1 = set(model_one_vars)
+    temp_set = [k for k in all_vars if k.name.startswith("generator")]
+    model_two_vars = [o for o in all_vars if o not in set_1]
+    '''print(len(all_vars))
+    print(len(model_one_vars))
+    print(len(temp_set))
+    print(len(model_two_vars))'''
+    #for i in model_one_vars :
+      #print(i)
+
+    ###################################################################
+    tf.train.Saver(model_two_vars).restore(sess, os.getcwd()+'/weights')
+    output = sess.run([ys_est, ys_res+PER_CHANNEL_MEANS],
+                      feed_dict={xs: imgs-PER_CHANNEL_MEANS})
+    plt.axis("off")
+    plt.imshow(output[0][0])
+    saveimg(output[0][0], 'output/IMG-%d-EnhanceNet.png' % i)
+    saveimg(recon_HR_img[i], 'output/IMG-%d-HR.png' % i)
+    saveimg(x_test[i], 'output/IMG-%d-LR.png' % i)
+    #halter = input("halted")
+    #saveimg(output[1][0], 'output/%s-Bicubic.png' % fne)
+    sess.close()
+    tf.reset_default_graph()
+

models/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "/content/drive/App/EnhanceNet/models/model.ckpt"
+all_model_checkpoint_paths: "/content/drive/App/EnhanceNet/models/model.ckpt"

models/last_model/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "/content/drive/App/EnhanceNet/models/model.ckpt"
+all_model_checkpoint_paths: "/content/drive/App/EnhanceNet/models/model.ckpt"