- Notifications
You must be signed in to change notification settings - Fork19
A deep learning based model to judge the AQ, Appearance Quotient, of faces. (For Chinese Young Girls Only)
NotificationsYou must be signed in to change notification settings
Entropy-xcy/RankFace
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
A deep learning based model to judge the AQ, Appearance Quotient, of faces. (For Chinese Young Girls Only)
Inspired byFace Rank
My Repository is just a reversion ofFace Rank. For more details check this fantastic repo.
ThisEssay along with its dataset gave me great help in modeling and handling training issues.
apt-get install python-dev python-pip -ygit clone https://github.com/Entropy-xcy/RankFacecd ./RankFacepip install -r requirements.txtapt-get install python-opencv# for macOS use 'brew install opencv'# for Windows try the installation tutorial from opencv official websitewget http://entropy-xcy.bid/faceRank.h5
python main.py girls.jpgHere is the output
It is highly recommended to train the model yourself. Some accuracy issues may happen if the platform you have is different from the trainer's.
rm ./faceRank.h5wget http://entropy-xcy.bid/dataset.zipunzip dataset.ziprm dataset.zip# You may change parameters in the script.python train.pyA basic webpage or POST API server build with kerasIt may still work for mobile platforms
pip install werkzeugpip install flask# make sure that you already successfully launched the demo before the next step# The default port is 5000, you may change it as you wish in the codepython API_server.py_________________________________________________________________Layer (type) Output Shape Param # =================================================================conv2d_1 (Conv2D) (None, 128, 128, 32) 896 _________________________________________________________________activation_1 (Activation) (None, 128, 128, 32) 0 _________________________________________________________________conv2d_2 (Conv2D) (None, 126, 126, 32) 9248 _________________________________________________________________activation_2 (Activation) (None, 126, 126, 32) 0 _________________________________________________________________max_pooling2d_1 (MaxPooling2 (None, 63, 63, 32) 0 _________________________________________________________________dropout_1 (Dropout) (None, 63, 63, 32) 0 _________________________________________________________________flatten_1 (Flatten) (None, 127008) 0 _________________________________________________________________dense_1 (Dense) (None, 128) 16257152 _________________________________________________________________activation_3 (Activation) (None, 128) 0 _________________________________________________________________dropout_2 (Dropout) (None, 128) 0 _________________________________________________________________dense_2 (Dense) (None, 1) 129 =================================================================Total params: 16,267,425Trainable params: 16,267,425Non-trainable params: 0_________________________________________________________________