The keras model is created by training SmallerVGGNet from scratch on around 2200 face images (~1100 for each class). Face region is cropped by applying face detection using cvlib on the images gathered from Google Images. It acheived around 96% training accuracy and ~90% validation accuracy. (20% of the dataset is used for validation)
Checkout the gender detection functionality implemented in cvlib which can be accessed through a single function call detect_gender().
- numpy
- opencv-python
- tensorflow
- keras
- requests
- progressbar
- cvlib
Install the required packages by executing the following command.
$ pip install -r requirements.txt
Note: Python 2.x is not supported
Make sure pip is linked to Python 3.x (pip -V will display this info).
If pip is linked to Python 2.7. Use pip3 instead.
pip3 can be installed using the command sudo apt-get install python3-pip
Using Python virtual environment is highly recommended.
$ python detect_gender.py -i <input_image>
$ python detect_gender_webcam.py
When you run the script for the first time, it will download the pre-trained model from this link and place it under pre-trained directory in the current path.
(If python command invokes default Python 2.7, use python3 instead)
You can download the dataset I gathered from Google Images from this link and train the network from scratch on your own if you are interested. You can add more images and play with the hyper parameters to experiment different ideas.
- scikit-learn
- matplotlib
Install them by typing pip install scikit-learn matplotlib
Start the training by running the command
$ python train.py -d <path-to-dataset>
(i.e) $ python train.py -d ~/Downloads/gender_dataset_face/
Depending on the hardware configuration of your system, the execution time will vary. On CPU, training will be slow. After the training, the model file will be saved in the current path as gender_detection.model.
If you have an Nvidia GPU, then you can install tensorflow-gpu package. It will make things run a lot faster.
If you are facing any difficulty, feel free to create a new issue or reach out on twitter @ponnusamy_arun .