Dealine: 26.02.2021
Please put your name here:
Name: Erlisa Kulla
- Fork the current repository
- Study the new framework-code of
- main.cpp
- Check that the code is running correctly: it should show the video stream from the web-camera of your laptop.
- Calculate average fps and print it to console every 2 seconds. Compare Debug and Release versions.
MacOS users may need to launch the application with the administrator rights, to grant access to the web-camera.
In both Release and Debug versions my console displays and average of 30 fps.
- Read the OpenCV documentation about Viola-Jones face detector: Cascade Classifier
- Implement face detection for the video stream from the web-camera using the
cv::CascadeClassifierclass. - Measure the FPS one more time. How FPS changed after incorporating the face detection into the framework?
Please do not copy-paste the example code from the OpenCV documentation, but try to understand the example code and implement the solution to the problem by yourself.
Compared to the initial calculation of the fps, it is way lower. I implemented a variable scale which helps in making scaling the video output.
The higher this value is, the less lag there will be, therefore the higher the fps will be. At a scale 4, the fps reaches around 14-15.5 with slight variations, while with scale 3 it is even lower.
This is due to the implementing the face detection algorithm which takes a lot of time and processing power to compare frames to given samples and constantly produce an output.
Please submit the assignment by making a pull request. Important : Please make sure that
- No extra files are submitted (except those, which were mentioned in the assignment)
- The changes were made only in those files where you were asked to write your code
- The Continiouse Integration system (appVeyor) can build the submitted code
- The rendered images are also submitted in the folder "renders"