[GSoC] Face Aging with Age-cGAN

tensorflow_examples/models/age_cgan/README.md

@@ -0,0 +1,45 @@
+# Age-cGAN
+### Face aging - FaceApp
+---
+#### Dependencies
+```
+tensorflow==2.0.0b1
+absl_py==0.7.0
+numpy==1.16.4
+matplotlib==2.2.3
+scipy==1.1.0
+```
+#### Downloads
+- To download all the dependencies, simply execute 
+```
+pip install -r requirements.txt
+```
+- To download the Wiki Cropped Faces dataset, simply execute the `data_download.py` file
+```
+python data_download.py
+```
+#### Training
+There are 3 stages of training:
+1. Conditional GAN Training (Generator and Discriminator)
+2. Initial Latent Vector Approximation (Encoder)
+3. Latent vector optimization (Encoder and Generator)
+- The `model.py` file contains the code to run all 3 stages of training. It automatically stores the weights after the specified/default number of epochs have completed. Note that the weights will be stored at the same directory level as `model.py`.
+```
+python model.py
+```
+- An important distinction to note is that there are 3 bool variables, namely `TRAIN_GAN`, `TRAIN_ENCODER` and `TRAIN_ENC_GAN` wich respectively represent the 3 stages of training. They have been set to `True` by default. If the requirement arises such that the training be completed individually, set the respective variables to `False` and ensure the requirements are met before proceeding.
+---
+#### Architecture
+1. Encoder: Learns the reverse mapping of input face images and the latent vector
+2. FaceNet: Face recognition network
+3. Generator: Takes a hidden representation of the face and a condition vector to generate a face
+4. Discriminator: Distinguishes between real and fake images
+#
+- **Encoder**: 4 CNN blocks with BatchNorm and Activation layers followed by 2 FC layers
+- **Generator**: CNNs, Upsampling layers and dense layers
+- **Face Recognition**: Pretrained `Inception-ResNet-V2` or `ResNet-50`
+- **Discriminator**: CNN blocks with BatchNorm and Activation layers
+#
+#### Reference Papers
+1. Face Aging with Conditional Generative Adversarial Networks [[Arxiv Link](https://arxiv.org/abs/1702.01983)]
+2. Improved Techniques for Training GANs [[Arxiv Link](https://arxiv.org/pdf/1606.03498.pdf)]

tensorflow_examples/models/age_cgan/data_download.py

@@ -0,0 +1,56 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Download the data.
+
+Dataset Citation:
+@article{Rothe-IJCV-2016,
+  author = {Rasmus Rothe and Radu Timofte and Luc Van Gool},
+  title = {Deep expectation of real and apparent age from a single image without facial landmarks},
+  journal = {International Journal of Computer Vision (IJCV)},
+  year = {2016},
+  month = {July},
+}
+"""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import argparse
+import tensorflow as tf
+assert tf.__version__.startswith('2')
+
+ap = argparse.ArgumentParser()
+ap.add_argument('-dp', '--download_path', required=False, help='Download Path')
+args = vars(ap.parse_args())
+
+data_url = "https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/static/wiki_crop.tar"
+
+def download_data(download_path):
+	path_to_zip = tf.keras.utils.get_file(
+		'wiki_crop.tar', cache_subdir=download_path,
+		origin = data_url, extract=True)
+
+	path_to_folder = os.path.join(os.path.dirname(path_to_zip), '')
+
+	return path_to_folder
+
+if __name__ == '__main__':	
+	if args['download_path'] is not None:
+		path = download_data(args["download_path"])
+	else:
+		cur = os.getcwd()
+		path = download_data(cur)