-
-
Confusion Matrix
-
Facial Expression Recognition
-
Webpage
Inspiration
After exams do you wonder how other people feel about the exam? Yes, our model has the ability to give you a slight idea of how the person in the video might feel about the exam, by emotion recognition.
What it does
It takes in a video mp4 file (maybe of you taking an exam or checking your answer with friends) and evaluate each frame to collect your facial emotion information. Then conclude your overall emotion and infer your performance on an exam.
How we built it
The model first detects face and then recognize emotions from those faces. We process videos frame-by-frame to collect data. The face detection model is based on Meta's end-to-end object detection model. Then from fer2013 data set, we filter out some controversial training data and use a CNN to train our emotion recognition model. And finally with the models, we can predict emotions from human face photos.
Challenges we ran into
The training data set was really large and partial unfiltered, it took us a lot of time to select the good training data.
Accomplishments that we're proud of
Training a model and videos processing.
What we learned
Think carefully and logically before making a move.
What's next for GuessExamPerformance
Probably deception detection through maybe micro expression analysis.
Built With
- computer-vision
- detr
- facial-expression-recognition
- ml
- python
Log in or sign up for Devpost to join the conversation.