We have three different applications.
- python3
- opencv-python
- tensorflow
- pandas
- numpy
- matplot
- sklearn
- glob
- os
- time
- Depending on your OS (unix / windows), you will have to adjust the paths in the code.
- The images of the dataset are not included in our submission. The positives would need to be copied into the
\dataset\positives\directory, the negative images into\dataset\negatives\.
Detects faces and smiles using your primary webcam. Take a screenshot with s, quit the application with q.
from the root folder
python3 liveSmileDetection.py.
Detects smiles from the given dataset. It takes about 20sec (on our machines). It will show the confusion matrix at the end.
from the root folder
python3 detectSmiles.py.
A model is already trained (located in the \model directory). However if you wish to train it yourself, it can be trained. It might take about 5 minutes to run. A GPU is not necessary but recommended. The dataset splitting is called from the application. It will save a new model in the \model directory and it will show two graphs, one with the accuracy over the epochs and one with the loss.
from the root folder
python3 TensorFlow/trainModel.py.
This application will predict, if an image contains a smile or frown. It will show the image with the prediction for each class as a label.
from the root folder
python3 TensorFlow/predict.py.
If you wish to predict another image, adjust the path in line 28 to your desired image path.
predictImage("[imagePath]")
All files in the \misc directory are either utility files or miscellaneous files. The cascade classifiers are located in this directory aswell, but will be loaded from the cv2 path.