-Implemented in Python 3.6, OpenCV2 library, Tensorflow, Matplotlib, Scipy.
Description of developing application:
My thesis aim to segment motion blurred object in video.
-
- Implement GrabCut method for semi-automatic segmentation of object of interest.
-
- OSVOS - One shot video object Segmentation - state of the art in precision(79% for one labeled image, 86% for 3 labeled images).
-
- Compare solution of detection of motion blur features & Optical flow solutions & Alpha Matting solutions.
-
- Implement one of these solutions and compare with state-of-the-art results.
Links:
-Thesis: https://github.com/sarvasrobert/Diploma/blob/master/Diploma_thesis.pdf
-OSVOS (Tensorflow): https://github.com/scaelles/OSVOS-TensorFlow
-Python 3.6 Download: https://www.python.org/downloads/
-OpenCV2 & Matplotlib library: https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_setup/py_setup_in_windows/py_setup_in_windows.html
You can find whole application under diploma folder.
Application consist of main.py, main_support.py, classifier.py, OSVOS folder, KNN_Matting folder.
Screens:
Outputs and testing of OSVOS object segmentation from video based on 1 labeled frame.
1.Dog 100 iterations, res:800x450px
2.Dog 100 iterations, res:800x450px
3.Man 100 iterations, res:800x450px
