In the age of a global pandemic the entire industry has shifted to a work from home enviornment. However Student Examintions taken online are still a tricky problem to solve as there are still a lot of loop holes for students to use while giving online exams or sitting for online lectures.
The Student Anti-Cheat Tool helps reduce these problem by detecting students faces with face recognition for identification and Students onscreen time with cellphone cheating detection.
Note :- Cell phone Detection uses Yolo V4 for object detection and will impact performance.
Requirements : Create a virtual conda enviornment.
conda create --name StudentAntiCheatEnv --file requirements.txt
conda activate StudentAntiCheatEnv
Usage:
-
Place Students Face photo with their names as "Firstname Lastname.jpg" in the "Faces/" folder in project diretory for face recognition and identification.
-
Place the model file in correct folder i.e. models/yolo/yolov4.h5 Yolo V5 Model Download Link
-
Press 'Q' during rendering to abort.
- Similar Usage for webcamDemo.py. This will render your webcam feed live with object detection and face detection for testing.
python StudentAntiCheat.py --path "TestVideo.mp4" --name "FirstName LastName" --fps 12 --phone true --save --verbose
Output:
Requires Cuda and CuDNN along with Tensorflow GPU Else CPU inference for phone detection will be extreemely slow
TensorFlow version: 2.1.0
Eager execution: True
Keras version: 2.2.4-tf
Cuda version: 10.1
Cudnn version: 7.6
Num Physical GPUs Available: 1
Num Logical GPUs Available: 1
Note: Depending on certain Windows machines the face_recogntion library may not install correctly.
pip install cmake
pip install dlib
pip install face_recognition
If still having issue you have to build dlib with cmake.
Refrences : YoloV4 Tensorflow by https://github.com/RobotEdh/Yolov-4