Inspiration
During the challenging times of the Covid pandemic, many students experienced a sense of melancholy and lack of enthusiasm towards attending school. The main culprit behind this disengagement was often attributed to the lengthy and monotonous lessons delivered by teachers. Unfortunately, educators were unaware of the emotional state we were in. As a response to this issue, we embarked on developing a program specifically designed to provide teachers with valuable feedback on their students' emotional well-being.
What It Does
Our innovative program offers a seamless solution for educators and individuals alike. By simply uploading a video file of a recorded virtual meeting, our program initiates an advanced video processing application. As the transformed video plays, sophisticated facial recognition technology precisely detects and outlines individuals' faces, displaying their corresponding emotions in real-time. This powerful feature empowers teachers to effortlessly gauge the emotional experiences of their students, enabling them to discern whether the classroom environment evoked positive or negative sentiments. 7 Different emotions will be identified.
How We Built The Detector
The provided code initializes the necessary models and parameters for face detection and emotion classification. It captures video frames, detects faces using a Haar cascade classifier, and classifies the emotions of the detected faces using our trained model file. The code tracks the dominant emotion by calculating the mode of the last few predictions and overlays bounding boxes and emotion labels on the video frames. The processed frames are displayed in a window until the 'q' key is pressed, at which point the video capture is released, and the windows are closed.
How We Built The Model
The code loads training data from a directory, normalizes it, and defines a CNN model using the mini_XCEPTION function. The model is compiled with an optimizer, loss function, and metrics. Training is performed for one epoch using the training dataset and validated using the validation dataset. The trained model is saved for future use.
Log in or sign up for Devpost to join the conversation.