init: commit

.gitattributes

@@ -0,0 +1,4 @@
+model/*.h5 filter=lfs diff=lfs merge=lfs -text
+model/resnet.h5 filter=lfs diff=lfs merge=lfs -text
+model/vggnet5.h5 filter=lfs diff=lfs merge=lfs -text
+model/vggnet.h5 filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,128 @@
+###DATA###
+MNIST/
+
+###Python###
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.cache
+nosetests.xml
+coverage.xml
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+
+###PyCharm###
+
+# PyCharm
+# http://www.jetbrains.com/pycharm/webhelp/project.html
+.idea
+.iml
+
+
+###IPythonNotebook###
+
+# Temporary data
+.ipynb_checkpoints/
+
+
+###OSX###
+
+.DS_Store
+.AppleDouble
+.LSOverride
+
+# Icon must end with two \r
+Icon
+
+
+# Thumbnails
+._*
+
+# Files that might appear on external disk
+.Spotlight-V100
+.Trashes
+
+# Directories potentially created on remote AFP share
+.AppleDB
+.AppleDesktop
+Network Trash Folder
+Temporary Items
+.apdisk
+
+
+###Linux###
+
+*~
+
+# KDE directory preferences
+.directory
+
+
+###Windows###
+
+# Windows image file caches
+Thumbs.db
+ehthumbs.db
+
+# Folder config file
+Desktop.ini
+
+# Recycle Bin used on file shares
+$RECYCLE.BIN/
+
+# Windows Installer files
+*.cab
+*.msi
+*.msm
+*.msp
+
+# Windows shortcuts
+*.lnk

README.md

@@ -0,0 +1,71 @@
+[resnet]: images/resnet.png "ResNet Model"
+[vggnet]: images/vggnet.png "VggNet Model"
+[vggnet5]: images/vggnet5.png "VggNet5 Model"
+
+# MNIST Competition Tensorflow KR Group
+
+* MNIST competition submission files
+* Used Keras
+* [Model Architecture](#model-architectures)
+    * [ResNet](#resnet)
+    * [VggNet](#vggnet)
+    * [VggNet5](#vggnet5)
+
+## Performance
+
+|  **Model**  |                 **Description**                 | **Accuracy** |
+|:-----------:|:-----------------------------------------------:|:------------:|
+|   VGG-like  |             VGGNet-like but smaller             |    99.71%    |
+| Resnet-like |             ResNet-like but smaller             |    99.60%    |
+|   VGG-like  | VGGNet-like but even smaller than the first one |    99.63%    |
+|  **Final**  |          **Ensemble 3 models + Voting**         |  **99.80%**  |
+
+## Run
+
+#### Evaluation
+```bash
+python evaluation.py
+```
+
+#### Train
+```bash
+python resnet.py 10 # 10 epochs & resnet
+python vgg16.py 10 # 10 epochs & vgg
+python vgg5.py 10 # 10 epochs & vgg
+```
+
+## File descriptions
+```bash
+├── evaluation.py # evaluation.py
+├── images # model architectures
+│   ├── resnet.png
+│   ├── vggnet5.png
+│   └── vggnet.png
+├── MNIST # mnist data (not included in this repo)
+│   ├── t10k-images-idx3-ubyte.gz
+│   ├── t10k-labels-idx1-ubyte.gz
+│   ├── train-images-idx3-ubyte.gz
+│   └── train-labels-idx1-ubyte.gz
+├── model # model weights
+│   ├── resnet.h5
+│   ├── vggnet5.h5
+│   └── vggnet.h5
+├── model.py # base model interface
+├── README.md
+├── utils.py # helper functions
+├── resnet.py
+├── vgg16.py
+└── vgg5.py
+```
+
+
+## Model Architectures
+
+#### ResNet
+![resnet][resnet]
+
+#### VggNet
+![vggnet][vggnet]
+
+#### VggNet5
+![vggnet5][vggnet5]

evaluation.py

@@ -0,0 +1,66 @@
+import numpy as np
+
+from vgg16 import VGGNet
+from vgg5 import VGGNet5
+from resnet import ResNet50 as ResNet
+from utils import load_mnist
+
+
+def load_models():
+    """Load models """
+    model_list = []
+
+    model_list.append(VGGNet("model/vggnet.h5"))
+    model_list.append(ResNet("model/resnet.h5"))
+    model_list.append(VGGNet5("model/vggnet5.h5"))
+
+    return model_list
+
+
+def evaluate(prediction, true_labels):
+    """Return an accuracy
+
+    Parameters
+    ----------
+    prediction : 2-d array, shape (n_sample, n_classes)
+        Onehot encoded predicted array
+
+    true_labels : 2-d array, shape (n_sample, n_classes)
+        Onehot encoded true array
+
+    Returns
+    ----------
+    accuracy : float
+        Return an accuracy
+    """
+    pred = np.argmax(prediction, 1)
+    true = np.argmax(true_labels, 1)
+
+    equal = (pred == true)
+
+    return np.mean(equal)
+
+
+def main():
+    model_list = load_models()
+    _, _, (X_test, y_test) = load_mnist()
+
+    pred_list = []
+
+    for idx, model in enumerate(model_list):
+        pred = model.predict(X_test)
+        pred_list.append(pred)
+
+        # Check a single model accuracy
+        acc = evaluate(pred, y_test)
+        print(f"Model-{idx}: {acc:>.5%}")
+
+    pred_list = np.asarray(pred_list)
+    pred_mean = np.mean(pred_list, 0)
+
+    accuracy = evaluate(pred_mean, y_test)
+    print(f"Final Test Accuracy: {accuracy:>.5%}")
+
+
+if __name__ == '__main__':
+    main()