August 7th: visualization function for Loss and Accuracy added.
- Soo Woon (Shawn) Chung
- Jingwen (Rebecca) Du
- Jingxian (Phebe) Lan
- Zhongyuan (Jasper) Zhang
Here is a collection of code and datasets corresponding to our project "Gender Detection from Facial Images with and without Mask".
Due to size limit, The Google Shared Folder https://drive.google.com/drive/folders/14a38EqwCzzJNp3faOuxFNfnFgyEW-hPO?usp=sharing includes the following files:
Datasets: CelebA, Casia WebFace Facial, Google Crawled
Best Model's Weight :weights6.best.inc.male.hdf5
Otherwise, all required files are stored in the github repository.
Before training the model, Unzip datasets in to imgs/
folder.
- python==3.7
- tensorflow==2.0
- keras==2.3.1
- matplotlib==3.3.0
- opencv-python==4.3.0
The directory structure is as follows:
genderdetect/
|-> main.py/ * changing hyperparameters and set folder path
|-> load_attr.py * Load image info
|-> training.py * network structures, train/test functions
|-> generate.py * data preprocessing and loading
|-> plot.py * results visualization
We also provide a notebookfile 7. DLPROJECT2-TestCutout_5squares_size30.ipynb containing the entire data preprocessing, training, validation, testing steps for convenience.