#ace-Matting-using-Unet Text-Detection-Model Text Detection Model trained with Precision Score of 98% on Test Set of 200 images.
Potrait Matting Model with 90% accuracy build on fine tuning vgg_unet using https://github.com/switchablenorms/CelebAMask-HQ Dataset. This model is made using amazing package made by Divam Gupta : https://divamgupta.com. Link to his repo https://github.com/divamgupta/image-segmentation-keras
- create a viruatual environment and activate it
- install required libs
apt-get install -y libsm6 libxext6 libxrender-dev
pip install opencv-python
- install keras-segmenatation pacakge using
pip install keras-segmentation
-
Downlaod Celeb Dataset
-
Edit the path to the dataset in preprocessing.py and run it.
python preprocessing.py
- Run the traning.py file using
python traning.py
- Test the image using inference-image.py
Disclaimer - I do not own/clicked any of the images below.
- Face attritube changing like color of the hair,skin
- potrait regeneration using Gans, regenerating new attributes like long hair, beard, etc
- Traninig on dataset with lesser quality to apply matting on non-hd images